Support Vector Regression with SGD Training From Scratch Using C#

Posted on June 22, 2026 by jamesdmccaffrey

The goal of a machine learning regression problem is to predict a single numeric value. For example, a bank might want to predict the maximum safe load amount for a customer, based on age, account balance, annual income, and so on.

One of about a dozen common regression techniques is (kernel) support vector regression (SVR). One day before work, I realized that I had not implemented SVR from scratch, using C#, for a long time. So I figured I’d take a stab at it.

There are three main ways to train a support regression model: quadratic programming (QP) optimization, the sequential minimal optimization (SMO) algorithm, and stochastic sub-gradient descent (SGD). I used SGD, which is far by the simplest SVR training technique.

The output of my demo is:

Begin C# kernel support vector regression with SGD

Loading train (200) and test (40) data
Done

First three X predictors:
  -0.1660   0.4406  -0.9998  -0.3953  -0.7065
   0.0776  -0.1616   0.3704  -0.5911   0.7562
  -0.9452   0.3409  -0.1654   0.1174  -0.7192

First three target y:
  0.4840
  0.1568
  0.8054

Creating SVR object
Setting RBF gamma = 0.3000
Setting epsilon = 0.007500
Setting C = 1.00
Setting lrnRate = 0.0010
Setting maxEpochs = 5000
Setting tol = 0.000100
Done

Training SVR model using SGD
epoch =      0 MSE = 0.0430 acc = 0.1300
epoch =   1000 MSE = 0.0001 acc = 0.9850
epoch =   2000 MSE = 0.0001 acc = 0.9850
epoch =   3000 MSE = 0.0001 acc = 0.9800
epoch =   4000 MSE = 0.0001 acc = 0.9800
Done


Model alpha (weights):
 -0.9256  -0.0443  -0.0041  -0.6581  . . .   0.0049
  0.2983   0.4573  -0.0488   0.0956  . . .   0.1471
. . .
  0.0017   0.0023   0.3193   0.0063  . . .  -0.0078
 -0.0690   0.9895   . . .    0.3843
Model bias = 0.4030

Number supp vectors = 194

Evaluating model

Train acc (within 0.10) = 0.9900
Test acc (within 0.10) = 0.9250

Train MSE = 0.0001
Test MSE = 0.0001

Train R2 = 0.9985
Test R2 = 0.9949

Predicting for trainX[0]
Predicted y = 0.4922

End SVR with SGD training demo

The demo data is synthetic. It was generated by a 5-10-1 neural network with random weights and bias values. The idea here is that the synthetic data does have an underlying, but complex, non-linear structure which can be predicted.

During training, the SVR model assigns one weight value, into a vector called alpha, to each training item, plus a special weight called the bias. After training, the SVR model determined that 6 of the 200 alpha weights were very close to zero, and so those 6 alpha values and their 6 associated training items were removed. This left 194 weight values and training items, called the support vectors.

All of the parameter values must be determined by trial and error. The gamma parameter defines the RBF function that is used to measure the similarity between data item vectors. Larger values of gamma shrinks the radius of influence of individual training points. This tends to increase model accuracy at the expense of increased risk of model overfitting.

The epsilon value defines how close to correct a prediction must be to be considered a non-support vector. Larger values of epsilon create fewer support vectors.

The C value is used for model regularization, which prevents model alpha weights from becoming very large, which often leads to model overfitting. Larger values of C have a smaller regularization affect.

The lrnRate value controls how much alpha weight values change at each update during training. Larger values of lrnRate increase the speed of training, at the risk of jumping over good weight values.

The maxEpochs value controls how many iterations are performed during training. The effect of larger values of SVR maxEpochs can vary greatly.

The tol (“tolerance”) value controls pruning away training vectors to support vectors, by defining how to close to 0 an alpha weight value must be in order to be pruned away. Larger values of tol allow more alpha weights to be defined as zero, which reduces the number of support vectors.

The biggest weakness of support vector regression is the difficulty of tuning the hyperparameters. Small changes in parameter values can have extremely large changes in the model, and the hyperparameters interact in complex ways.

Support vector regression had a brief surge of popularity in the late 1990s and early 2000s. However, data scientists realized that the closely related kernel ridge regression (KRR) has several significant advantages over SVR, and so the use of SVR declined to the point where it is not used very much today.

SVR is more difficult to implement than KRR, SVR is much more difficult to tune than KRR (KRR can use true SGD, which is easier to tune than SVR sub-gradient descent), and SVR often gives slightly worse prediction accuracy than KRR (due mostly to the difficulty in parameter tuning). That said, there are some problem scenarios where kernel SVR is highly effective.


One of the advantages of implementing a machine learning regression system from scratch is that it gives you a full explanation and understanding of exactly how the system works.

I’m a big fan of science fiction movies from the 1950s and 1960s. I especially like scenes where a scientist explains what the threat is.

Left: In “Crack in the World” (1965), a scientist explains how his project that will drill a hole into the Earth’s magma core (using a nuclear missile), to gain access to unlimited power. Well, based on the title of this movie, you can correctly guess that this idea did not work out very well. My grade for this movie = B.

Right: In “Gorath” (1962), a scientist explains that a runaway star named Gorath is on a collision course with Earth. All governments unite to build gigantic rocket thrusters at the South Pole to move Earth out of the way, and then back again. The plan succeeds. My grade for this movie = B.

Demo program. Replace “lt” (less than), “gt”, “lte”, “gte” with Boolean operator symbols (my lame blog editor often chokes on symbols).

using System;
using System.IO;
using System.Collections.Generic;

// kernel SVR with SGD training
// hard-wired RBF kernel function

namespace SupportVectorRegressionSGD
{
  internal class SupportVectorRegressionSGDProgram
  {
    static void Main(string[] args)
    {
      Console.WriteLine("\nBegin C# kernel support " +
        "vector regression with SGD ");

      Console.WriteLine("\nLoading train (200) and" +
        " test (40) data ");
      string trainFile =
        "..\\..\\..\\Data\\synthetic_train_200.txt";
      double[][] trainX =
        Utils.MatLoad(trainFile,
        new int[] { 0, 1, 2, 3, 4 }, ',', "#");
      double[] trainY =
        Utils.MatToVec(Utils.MatLoad(trainFile,
        new int[] { 5 }, ',', "#"));

      string testFile =
        "..\\..\\..\\Data\\synthetic_test_40.txt";
      double[][] testX =
        Utils.MatLoad(testFile,
        new int[] { 0, 1, 2, 3, 4 }, ',', "#");
      double[] testY =
        Utils.MatToVec(Utils.MatLoad(testFile,
        new int[] { 5 }, ',', "#"));
      Console.WriteLine("Done ");

      Console.WriteLine("\nFirst three X predictors: ");
      for (int i = 0; i "lt" 3; ++i)
        Utils.VecShow(trainX[i], 4, 9);
      Console.WriteLine("\nFirst three target y: ");
      for (int i = 0; i "lt" 3; ++i)
        Console.WriteLine(trainY[i].ToString("F4").
          PadLeft(8));

      Console.WriteLine("\nCreating SVR object");
      double gamma = 0.30;    // RBF param
      double epsilon = 0.0075;
      double C = 1.0;
      double lrnRate = 0.001;
      int maxEpochs = 5000;
      double tol = 1.0e-4;

      Console.WriteLine("Setting RBF gamma = " +
        gamma.ToString("F4"));
      Console.WriteLine("Setting epsilon = " +
        epsilon.ToString("F6"));
      Console.WriteLine("Setting C = " +
        C.ToString("F2"));
      Console.WriteLine("Setting lrnRate = " +
        lrnRate.ToString("F4"));
      Console.WriteLine("Setting maxEpochs = " +
        maxEpochs);
      Console.WriteLine("Setting tol = " +
        tol.ToString("F6"));

      SVR model = new SVR(gamma, epsilon, C,
        lrnRate, maxEpochs, tol, seed: 0);
      Console.WriteLine("Done ");

      Console.WriteLine("\nTraining SVR model using" +
        " SGD ");
      model.Train(trainX, trainY);
      Console.WriteLine("Done ");
      Console.WriteLine("\nModel alpha (weights): ");
      Utils.VecShow(model.alpha, 4, 9);
      Console.WriteLine("\nModel bias = " + 
        model.b.ToString("F4"));

      Console.WriteLine("\nNumber supp vectors = " +
        model.suppX.Length);

      Console.WriteLine("\nEvaluating model ");
      double trainAcc =
        model.Accuracy(trainX, trainY, 0.10);
      double testAcc =
        model.Accuracy(testX, testY, 0.10);

      Console.WriteLine("\nTrain acc (within 0.10) = " +
        trainAcc.ToString("F4"));
      Console.WriteLine("Test acc (within 0.10) = " +
        testAcc.ToString("F4"));

      double trainMSE = model.MSE(trainX, trainY);
      double testMSE = model.MSE(testX, testY);

      Console.WriteLine("\nTrain MSE = " +
        trainMSE.ToString("F4"));
      Console.WriteLine("Test MSE = " +
        testMSE.ToString("F4"));

      double trainR2 = model.R2(trainX, trainY);
      double testR2 = model.R2(testX, testY);

      Console.WriteLine("\nTrain R2 = " +
        trainR2.ToString("F4"));
      Console.WriteLine("Test R2 = " +
        testR2.ToString("F4"));

      double[] x = trainX[0];
      Console.WriteLine("\nPredicting for trainX[0] ");
      double y = model.Predict(x);
      Console.WriteLine("Predicted y = " +
        y.ToString("F4"));

      Console.WriteLine("\nEnd SVR with SGD training demo ");
      Console.ReadLine();
    } // Main()

  } // class Program

  // ========================================================

  public class SVR
  {
    public double gamma;  // for RBF kernel
    public double epsilon;
    public double C; // weight regularization
    public double[][] suppX;  // needed for pred
    public double[] suppY;
    public double[] alpha;  // one per trainX item
    public double b;   // bias
    public double lrnRate;  // for SGD training
    public int maxEpochs;
    public double tol;
    public Random rnd;

    // ------------------------------------------------------

    public SVR(double gamma, double epsilon, double C,
      double lrnRate, int maxEpochs, double tol,
      int seed = 0)
    {
      this.gamma = gamma;
      this.epsilon = epsilon;
      this.C = C;
      this.suppX = new double[0][]; // compiler happy
      this.suppY = new double[0];
      this.lrnRate = lrnRate;
      this.maxEpochs = maxEpochs;
      this.tol = tol;
      this.alpha = new double[0];
      this.b = 0.0;
      this.rnd = new Random(seed);  // shuffle train order
    } // ctor

    // ------------------------------------------------------

    public void Train(double[][] trainX, double[] trainY)
    {
      this.suppX = trainX;
      this.suppY = trainY;
      int n = trainX.Length;

      // init weights
      this.alpha = new double[n];
      double lo = -0.01; double hi = 0.01; // not needed
      for (int i = 0; i "lt" n; ++i)
        this.alpha[i] =
          (hi - lo) * this.rnd.NextDouble() + lo;
      this.b = 0.0;

      // precompute all rbf values to K for fast train
      // not feasible for huge datasets
      double[][] K = this.MakeK(trainX);

      // set up indices for random order SGD training
      int[] indices = Utils.VecRange(n); // 0, 1, 2, ..
      double lamda = 1.0 / this.C;
      int progressFreq = (int)(this.maxEpochs / 5);

      // main sub-gradient processing loop
      for (int epoch = 0; epoch "lt" this.maxEpochs; ++epoch)
      {
        this.Shuffle(indices);
        for (int i = 0; i "lt" indices.Length; ++i)
        {
          int idx = indices[i];
          double predY = 0.0;
          for (int j = 0; j "lt" this.alpha.Length; ++j)
            predY += this.alpha[j] * K[idx][j]; // fast
          predY += this.b;
          double error = predY - trainY[idx];

          double gradLoss;
          bool insideTube = false;
          if (error "gt" this.epsilon)
            gradLoss = 1.0;
          else if (error "lt" -this.epsilon)
            gradLoss = -1.0;
          else
          {
            gradLoss = 0.0;
            insideTube = true;
          }

          // local kernel regularization gradient
          double gradReg = this.alpha[idx] * K[idx][idx];

          //  decoupled updates to the active index
          this.alpha[idx] -= this.lrnRate *
            (lamda * gradReg + gradLoss);
          this.b -= this.lrnRate * gradLoss;

          // force tiny weights to 0
          if (insideTube == true &&
            Math.Abs(this.alpha[idx]) "lt" this.tol)
            this.alpha[idx] = 0.0;

          // in-loop clip to bound updates mid-flight
          if (this.alpha[idx] "lt" -this.C)
            this.alpha[idx] = -this.C;
          else if (this.alpha[idx] "gt" this.C)
            this.alpha[idx] = this.C;

        } // each item

        // show training progress every few epochs
        if (epoch % progressFreq == 0)
        {
          double mse =
            this.MSE(trainX, trainY);
          double acc =
            this.Accuracy(trainX, trainY, 0.10);
          string s1 = "epoch = " +
            epoch.ToString().PadLeft(6);
          string s2 = " MSE = " +
            mse.ToString("F4");
          string s3 = " acc = " + acc.ToString("F4");
          Console.WriteLine(s1 + s2 + s3);
        }

      } // each epoch

      // final global clip
      for (int i = 0; i "lt" n; ++i)
      {
        if (this.alpha[i] "lt" -this.C)
          this.alpha[i] = -this.C;
        else if (this.alpha[i] "gt" this.C)
          this.alpha[i] = this.C;
      }

      // prune: store only explicit support vectors
      List"lt"int"gt" svLst = new List"lt"int"gt"();
      for (int i = 0; i "lt" this.alpha.Length; ++i)
      {
        if (Math.Abs(this.alpha[i]) "gt" 1.0e-5)
          svLst.Add(i);
      }
      int[] svMask = svLst.ToArray();

      this.suppX = Utils.MatSelectRows(trainX, svMask);
      this.suppY = Utils.VecSelectItems(trainY, svMask);
      this.alpha = Utils.VecSelectItems(this.alpha, svMask);

      return;  // all done
    } // Train

    // ------------------------------------------------------

    private void Shuffle(int[] indices)
    {
      // Fisher-Yates helper for Train()
      for (int i = 0; i "lt" indices.Length; ++i)
      {
        int ri = this.rnd.Next(i, indices.Length);
        int tmp = indices[i];
        indices[i] = indices[ri];
        indices[ri] = tmp;
      }
    } // Shuffle

    // ------------------------------------------------------

    private double RBF(double[] v1, double[] v2)
    {
      int n = v1.Length;
      double sum = 0.0;
      for (int i = 0; i "lt" n; ++i)
      {
        double d = v1[i] - v2[i];
        sum += d * d;
      }
      double result = Math.Exp(-1 * this.gamma * sum);
      return result;
    }

    // ------------------------------------------------------

    private double[][] MakeK(double[][] X)
    {
      // Kernel-Gram matrix helper for Train()
      // pre-compute all similarities, to avoid re-computes
      int n = X.Length;
      double[][] result = Utils.MatMake(n, n);
      for (int i = 0; i "lt" n; ++i)
        for (int j = 0; j "lt" n; ++j)
          result[i][j] = this.RBF(X[i], X[j]);
      return result;
    }

    // ------------------------------------------------------

    public double Predict(double[] x)
    {
      int n = this.suppX.Length;
      double sum = 0.0;
      for (int i = 0; i "lt" n; ++i)
      {
        double[] xx = this.suppX[i];
        double k = this.RBF(x, xx);
        sum += this.alpha[i] * k;
      }
      return sum + this.b;
    }

    // ------------------------------------------------------

    public double Accuracy(double[][] dataX,
      double[] dataY, double pctClose)
    {
      int numCorrect = 0; int numWrong = 0;
      int n = dataX.Length;
      for (int i = 0; i "lt" n; ++i)
      {
        double[] x = dataX[i];
        double actualY = dataY[i];
        double predY = this.Predict(x);
        if (Math.Abs(actualY - predY) "lt"
          Math.Abs(actualY * pctClose))
          ++numCorrect;
        else
          ++numWrong;
      }
      return (numCorrect * 1.0) / n;
    }

    // ------------------------------------------------------

    public double MSE(double[][] dataX, double[] dataY)
    {
      double sum = 0.0;
      int n = dataX.Length;
      for (int i = 0; i "lt" n; ++i)
      {
        double[] x = dataX[i];
        double actualY = dataY[i];
        double predY = this.Predict(x);
        sum += (actualY - predY) * (actualY - predY);
      }
      return sum / n;
    }

    // ------------------------------------------------------

    public double R2(double[][] dataX, double[] dataY)
    {
      // coefficient of determination
      int n = dataX.Length;
      double ssRes = 0.0; double ssTot = 0.0;
      double meanY = Utils.VecMean(dataY);

      for (int i = 0; i "lt" n; ++i)
      {
        double[] x = dataX[i];
        double actualY = dataY[i];
        double predY = this.Predict(x);
        ssRes += (actualY - predY) * (actualY - predY);
        ssTot += (actualY - meanY) * (actualY - meanY);
      }
      double result = 1.0 - (ssRes / ssTot);
      return result;
    }

  } // class KRR

  // ========================================================

  public class Utils
  {
    // ------------------------------------------------------

    public static double[][] MatLoad(string fn,
      int[] usecols, char sep, string comment)
    {
      List"lt"double[]"gt" result =
        new List"lt"double[]"gt"();
      string line = "";
      FileStream ifs = new FileStream(fn, FileMode.Open);
      StreamReader sr = new StreamReader(ifs);
      while ((line = sr.ReadLine()) != null)
      {
        if (line.StartsWith(comment) == true)
          continue;
        string[] tokens = line.Split(sep);
        List"lt"double"gt" lst = new List"lt"double"gt"();
        for (int j = 0; j "lt" usecols.Length; ++j)
          lst.Add(double.Parse(tokens[usecols[j]]));
        double[] row = lst.ToArray();
        result.Add(row);
      }
      sr.Close(); ifs.Close();
      return result.ToArray();
    }

    // ------------------------------------------------------

    public static double[] MatToVec(double[][] X)
    {
      int nRows = X.Length;
      int nCols = X[0].Length;
      double[] result = new double[nRows * nCols];
      int k = 0;
      for (int i = 0; i "lt" nRows; ++i)
        for (int j = 0; j "lt" nCols; ++j)
          result[k++] = X[i][j];
      return result;
    }

    // ------------------------------------------------------

    //public static double[] MatGetColumn(double[][] X,
    //  int col)
    //{
    //  int n = X.Length;
    //  double[] result = new double[n];
    //  for (int i = 0; i "lt" n; ++i)
    //    result[i] = X[i][col];
    //  return result;
    //}

    // ------------------------------------------------------

    public static double[][] MatSelectRows(double[][] X,
      int[] rows)
    {
      int nRowsSrc = X.Length;
      int nColsSrc = X[0].Length;
      int n = rows.Length;
      double[][] result = MatMake(n, nColsSrc);

      for (int i = 0; i "lt" n; ++i) // i pts into result
      {
        int srcRow = rows[i];
        for (int j = 0; j "lt" nColsSrc; ++j)
        {
          result[i][j] = X[srcRow][j];
        }
      }
      return result;
    }

    // ------------------------------------------------------

    public static double[][] MatMake(int nRows, int nCols)
    {
      double[][] result = new double[nRows][];
      for (int i = 0; i "lt" nRows; ++i)
        result[i] = new double[nCols];
      return result;
    }

    // ------------------------------------------------------

    public static double VecMean(double[] vec)
    {
      int n = vec.Length;
      double sum = 0.0;
      for (int i = 0; i "lt" n; ++i)
        sum += vec[i];
      double result = sum / n;
      return result;
    }

    // ------------------------------------------------------

    public static int[] VecRange(int n)
    {
      int[] result = new int[n];
      for (int i = 0; i "lt" n; ++i)
        result[i] = i;
      return result;
    }

    // ------------------------------------------------------

    //public static double VecDot(double[] v1, double[] v2)
    //{
    //  int n = v1.Length;
    //  double sum = 0.0;
    //  for (int i = 0; i "lt" n; ++i)
    //    sum += v1[i] * v2[i];
    //  return sum;
    //}

    // ------------------------------------------------------

    public static double[] VecSelectItems(double[] vec,
      int[] idxs)
    {
      int n = idxs.Length;
      double[] result = new double[n];
      for (int i = 0; i "lt" n; ++i)
      {
        result[i] = vec[idxs[i]];
      }
      return result;
    }

    // ------------------------------------------------------

    public static void VecShow(int[] vec, int wid)
    {
      for (int i = 0; i "lt" vec.Length; ++i)
        Console.Write(vec[i].ToString().PadLeft(wid));
      Console.WriteLine("");
    }

    // ------------------------------------------------------

    public static void MatShow(double[][] m, int dec,
      int wid)
    {
      int nRows = m.Length; int nCols = m[0].Length;
      double small = 1.0 / Math.Pow(10, dec);
      for (int i = 0; i "lt" nRows; ++i)
      {
        for (int j = 0; j "lt" nCols; ++j)
        {
          double v = m[i][j];
          if (Math.Abs(v) "lt" small) v = 0.0;
          Console.Write(v.ToString("F" + dec).
            PadLeft(wid));
        }
        Console.WriteLine("");
      }
    }

    // ------------------------------------------------------

    public static void VecShow(double[] vec, int dec,
      int wid)
    {
      for (int i = 0; i "lt" vec.Length; ++i)
        Console.Write(vec[i].ToString("F" + dec).PadLeft(wid));
      Console.WriteLine("");
    }

    // ------------------------------------------------------

  } // class Utils

  // ========================================================

} // ns

Training data:

# synthetic_train_200.txt
#
-0.1660,  0.4406, -0.9998, -0.3953, -0.7065,  0.4840
 0.0776, -0.1616,  0.3704, -0.5911,  0.7562,  0.1568
-0.9452,  0.3409, -0.1654,  0.1174, -0.7192,  0.8054
 0.9365, -0.3732,  0.3846,  0.7528,  0.7892,  0.1345
-0.8299, -0.9219, -0.6603,  0.7563, -0.8033,  0.7955
 0.0663,  0.3838, -0.3690,  0.3730,  0.6693,  0.3206
-0.9634,  0.5003,  0.9777,  0.4963, -0.4391,  0.7377
-0.1042,  0.8172, -0.4128, -0.4244, -0.7399,  0.4801
-0.9613,  0.3577, -0.5767, -0.4689, -0.0169,  0.6861
-0.7065,  0.1786,  0.3995, -0.7953, -0.1719,  0.5569
 0.3888, -0.1716, -0.9001,  0.0718,  0.3276,  0.2500
 0.1731,  0.8068, -0.7251, -0.7214,  0.6148,  0.3297
-0.2046, -0.6693,  0.8550, -0.3045,  0.5016,  0.2129
 0.2473,  0.5019, -0.3022, -0.4601,  0.7918,  0.2613
-0.1438,  0.9297,  0.3269,  0.2434, -0.7705,  0.5171
 0.1568, -0.1837, -0.5259,  0.8068,  0.1474,  0.3307
-0.9943,  0.2343, -0.3467,  0.0541,  0.7719,  0.5581
 0.2467, -0.9684,  0.8589,  0.3818,  0.9946,  0.1092
-0.6553, -0.7257,  0.8652,  0.3936, -0.8680,  0.7018
 0.8460,  0.4230, -0.7515, -0.9602, -0.9476,  0.1996
-0.9434, -0.5076,  0.7201,  0.0777,  0.1056,  0.5664
 0.9392,  0.1221, -0.9627,  0.6013, -0.5341,  0.1533
 0.6142, -0.2243,  0.7271,  0.4942,  0.1125,  0.1661
 0.4260,  0.1194, -0.9749, -0.8561,  0.9346,  0.2230
 0.1362, -0.5934, -0.4953,  0.4877, -0.6091,  0.3810
 0.6937, -0.5203, -0.0125,  0.2399,  0.6580,  0.1460
-0.6864, -0.9628, -0.8600, -0.0273,  0.2127,  0.5387
 0.9772,  0.1595, -0.2397,  0.1019,  0.4907,  0.1611
 0.3385, -0.4702, -0.8673, -0.2598,  0.2594,  0.2270
-0.8669, -0.4794,  0.6095, -0.6131,  0.2789,  0.4700
 0.0493,  0.8496, -0.4734, -0.8681,  0.4701,  0.3516
 0.8639, -0.9721, -0.5313,  0.2336,  0.8980,  0.1412
 0.9004,  0.1133,  0.8312,  0.2831, -0.2200,  0.1782
 0.0991,  0.8524,  0.8375, -0.2102,  0.9265,  0.2150
-0.6521, -0.7473, -0.7298,  0.0113, -0.9570,  0.7422
 0.6190, -0.3105,  0.8802,  0.1640,  0.7577,  0.1056
 0.6895,  0.8108, -0.0802,  0.0927,  0.5972,  0.2214
 0.1982, -0.9689,  0.1870, -0.1326,  0.6147,  0.1310
-0.3695,  0.7858,  0.1557, -0.6320,  0.5759,  0.3773
-0.1596,  0.3581,  0.8372, -0.9992,  0.9535,  0.2071
-0.2468,  0.9476,  0.2094,  0.6577,  0.1494,  0.4132
 0.1737,  0.5000,  0.7166,  0.5102,  0.3961,  0.2611
 0.7290, -0.3546,  0.3416, -0.0983, -0.2358,  0.1332
-0.3652,  0.2438, -0.1395,  0.9476,  0.3556,  0.4170
-0.6029, -0.1466, -0.3133,  0.5953,  0.7600,  0.4334
-0.4596, -0.4953,  0.7098,  0.0554,  0.6043,  0.2775
 0.1450,  0.4663,  0.0380,  0.5418,  0.1377,  0.2931
-0.8636, -0.2442, -0.8407,  0.9656, -0.6368,  0.7429
 0.6237,  0.7499,  0.3768,  0.1390, -0.6781,  0.2185
-0.5499,  0.1850, -0.3755,  0.8326,  0.8193,  0.4399
-0.4858, -0.7782, -0.6141, -0.0008,  0.4572,  0.4197
 0.7033, -0.1683,  0.2334, -0.5327, -0.7961,  0.1776
 0.0317, -0.0457, -0.6947,  0.2436,  0.0880,  0.3345
 0.5031, -0.5559,  0.0387,  0.5706, -0.9553,  0.3107
-0.3513,  0.7458,  0.6894,  0.0769,  0.7332,  0.3170
 0.2205,  0.5992, -0.9309,  0.5405,  0.4635,  0.3532
-0.4806, -0.4859,  0.2646, -0.3094,  0.5932,  0.3202
 0.9809, -0.3995, -0.7140,  0.8026,  0.0831,  0.1600
 0.9495,  0.2732,  0.9878,  0.0921,  0.0529,  0.1289
-0.9476, -0.6792,  0.4913, -0.9392, -0.2669,  0.5966
 0.7247,  0.3854,  0.3819, -0.6227, -0.1162,  0.1550
-0.5922, -0.5045, -0.4757,  0.5003, -0.0860,  0.5863
-0.8861,  0.0170, -0.5761,  0.5972, -0.4053,  0.7301
 0.6877, -0.2380,  0.4997,  0.0223,  0.0819,  0.1404
 0.9189,  0.6079, -0.9354,  0.4188, -0.0700,  0.1907
-0.1428, -0.7820,  0.2676,  0.6059,  0.3936,  0.2790
 0.5324, -0.3151,  0.6917, -0.1425,  0.6480,  0.1071
-0.8432, -0.9633, -0.8666, -0.0828, -0.7733,  0.7784
-0.9444,  0.5097, -0.2103,  0.4939, -0.0952,  0.6787
-0.0520,  0.6063, -0.1952,  0.8094, -0.9259,  0.4836
 0.5477, -0.7487,  0.2370, -0.9793,  0.0773,  0.1241
 0.2450,  0.8116,  0.9799,  0.4222,  0.4636,  0.2355
 0.8186, -0.1983, -0.5003, -0.6531, -0.7611,  0.1511
-0.4714,  0.6382, -0.3788,  0.9648, -0.4667,  0.5950
 0.0673, -0.3711,  0.8215, -0.2669, -0.1328,  0.2677
-0.9381,  0.4338,  0.7820, -0.9454,  0.0441,  0.5518
-0.3480,  0.7190,  0.1170,  0.3805, -0.0943,  0.4724
-0.9813,  0.1535, -0.3771,  0.0345,  0.8328,  0.5438
-0.1471, -0.5052, -0.2574,  0.8637,  0.8737,  0.3042
-0.5454, -0.3712, -0.6505,  0.2142, -0.1728,  0.5783
 0.6327, -0.6297,  0.4038, -0.5193,  0.1484,  0.1153
-0.5424,  0.3282, -0.0055,  0.0380, -0.6506,  0.6613
 0.1414,  0.9935,  0.6337,  0.1887,  0.9520,  0.2540
-0.9351, -0.8128, -0.8693, -0.0965, -0.2491,  0.7353
 0.9507, -0.6640,  0.9456,  0.5349,  0.6485,  0.1059
-0.0462, -0.9737, -0.2940, -0.0159,  0.4602,  0.2606
-0.0627, -0.0852, -0.7247, -0.9782,  0.5166,  0.2977
 0.0478,  0.5098, -0.0723, -0.7504, -0.3750,  0.3335
 0.0090,  0.3477,  0.5403, -0.7393, -0.9542,  0.4415
-0.9748,  0.3449,  0.3736, -0.1015,  0.8296,  0.4358
 0.2887, -0.9895, -0.0311,  0.7186,  0.6608,  0.2057
 0.1570, -0.4518,  0.1211,  0.3435, -0.2951,  0.3244
 0.7117, -0.6099,  0.4946, -0.4208,  0.5476,  0.1096
-0.2929, -0.5726,  0.5346, -0.3827,  0.4665,  0.2465
 0.4889, -0.5572, -0.5718, -0.6021, -0.7150,  0.2163
-0.7782,  0.3491,  0.5996, -0.8389, -0.5366,  0.6516
-0.5847,  0.8347,  0.4226,  0.1078, -0.3910,  0.6134
 0.8469,  0.4121, -0.0439, -0.7476,  0.9521,  0.1571
-0.6803, -0.5948, -0.1376, -0.1916, -0.7065,  0.7156
 0.2878,  0.5086, -0.5785,  0.2019,  0.4979,  0.2980
 0.2764,  0.1943, -0.4090,  0.4632,  0.8906,  0.2960
-0.8877,  0.6705, -0.6155, -0.2098, -0.3998,  0.7107
-0.8398,  0.8093, -0.2597,  0.0614, -0.0118,  0.6502
-0.8476,  0.0158, -0.4769, -0.2859, -0.7839,  0.7715
 0.5751, -0.7868,  0.9714, -0.6457,  0.1448,  0.1175
 0.4802, -0.7001,  0.1022, -0.5668,  0.5184,  0.1090
 0.4458, -0.6469,  0.7239, -0.9604,  0.7205,  0.0779
 0.5175,  0.4339,  0.9747, -0.4438, -0.9924,  0.2879
 0.8678,  0.7158,  0.4577,  0.0334,  0.4139,  0.1678
 0.5406,  0.5012,  0.2264, -0.1963,  0.3946,  0.2088
-0.9938,  0.5498,  0.7928, -0.5214, -0.7585,  0.7687
 0.7661,  0.0863, -0.4266, -0.7233, -0.4197,  0.1466
 0.2277, -0.3517, -0.0853, -0.1118,  0.6563,  0.1767
 0.3499, -0.5570, -0.0655, -0.3705,  0.2537,  0.1632
 0.7547, -0.1046,  0.5689, -0.0861,  0.3125,  0.1257
 0.8186,  0.2110,  0.5335,  0.0094, -0.0039,  0.1391
 0.6858, -0.8644,  0.1465,  0.8855,  0.0357,  0.1845
-0.4967,  0.4015,  0.0805,  0.8977,  0.2487,  0.4663
 0.6760, -0.9841,  0.9787, -0.8446, -0.3557,  0.1509
-0.1203, -0.4885,  0.6054, -0.0443, -0.7313,  0.4854
 0.8557,  0.7919, -0.0169,  0.7134, -0.1628,  0.2002
 0.0115, -0.6209,  0.9300, -0.4116, -0.7931,  0.4052
-0.7114, -0.9718,  0.4319,  0.1290,  0.5892,  0.3661
 0.3915,  0.5557, -0.1870,  0.2955, -0.6404,  0.2954
-0.3564, -0.6548, -0.1827, -0.5172, -0.1862,  0.4622
 0.2392, -0.4959,  0.5857, -0.1341, -0.2850,  0.2470
-0.3394,  0.3947, -0.4627,  0.6166, -0.4094,  0.5325
 0.7107,  0.7768, -0.6312,  0.1707,  0.7964,  0.2757
-0.1078,  0.8437, -0.4420,  0.2177,  0.3649,  0.4028
-0.3139,  0.5595, -0.6505, -0.3161, -0.7108,  0.5546
 0.4335,  0.3986,  0.3770, -0.4932,  0.3847,  0.1810
-0.2562, -0.2894, -0.8847,  0.2633,  0.4146,  0.4036
 0.2272,  0.2966, -0.6601, -0.7011,  0.0284,  0.2778
-0.0743, -0.1421, -0.0054, -0.6770, -0.3151,  0.3597
-0.4762,  0.6891,  0.6007, -0.1467,  0.2140,  0.4266
-0.4061,  0.7193,  0.3432,  0.2669, -0.7505,  0.6147
-0.0588,  0.9731,  0.8966,  0.2902, -0.6966,  0.4955
-0.0627, -0.1439,  0.1985,  0.6999,  0.5022,  0.3077
 0.1587,  0.8494, -0.8705,  0.9827, -0.8940,  0.4263
-0.7850,  0.2473, -0.9040, -0.4308, -0.8779,  0.7199
 0.4070,  0.3369, -0.2428, -0.6236,  0.4940,  0.2215
-0.0242,  0.0513, -0.9430,  0.2885, -0.2987,  0.3947
-0.5416, -0.1322, -0.2351, -0.0604,  0.9590,  0.3683
 0.1055,  0.7783, -0.2901, -0.5090,  0.8220,  0.2984
-0.9129,  0.9015,  0.1128, -0.2473,  0.9901,  0.4776
-0.9378,  0.1424, -0.6391,  0.2619,  0.9618,  0.5368
 0.7498, -0.0963,  0.4169,  0.5549, -0.0103,  0.1614
-0.2612, -0.7156,  0.4538, -0.0460, -0.1022,  0.3717
 0.7720,  0.0552, -0.1818, -0.4622, -0.8560,  0.1685
-0.4177,  0.0070,  0.9319, -0.7812,  0.3461,  0.3052
-0.0001,  0.5542, -0.7128, -0.8336, -0.2016,  0.3803
 0.5356, -0.4194, -0.5662, -0.9666, -0.2027,  0.1776
-0.2378,  0.3187, -0.8582, -0.6948, -0.9668,  0.5474
-0.1947, -0.3579,  0.1158,  0.9869,  0.6690,  0.2992
 0.3992,  0.8365, -0.9205, -0.8593, -0.0520,  0.3154
-0.0209,  0.0793,  0.7905, -0.1067,  0.7541,  0.1864
-0.4928, -0.4524, -0.3433,  0.0951, -0.5597,  0.6261
-0.8118,  0.7404, -0.5263, -0.2280,  0.1431,  0.6349
 0.0516, -0.8480,  0.7483,  0.9023,  0.6250,  0.1959
-0.3212,  0.1093,  0.9488, -0.3766,  0.3376,  0.2735
-0.3481,  0.5490, -0.3484,  0.7797,  0.5034,  0.4379
-0.5785, -0.9170, -0.3563, -0.9258,  0.3877,  0.4121
 0.3407, -0.1391,  0.5356,  0.0720, -0.9203,  0.3458
-0.3287, -0.8954,  0.2102,  0.0241,  0.2349,  0.3247
-0.1353,  0.6954, -0.0919, -0.9692,  0.7461,  0.3338
 0.9036, -0.8982, -0.5299, -0.8733, -0.1567,  0.1187
 0.7277, -0.8368, -0.0538, -0.7489,  0.5458,  0.0830
 0.9049,  0.8878,  0.2279,  0.9470, -0.3103,  0.2194
 0.7957, -0.1308, -0.5284,  0.8817,  0.3684,  0.2172
 0.4647, -0.4931,  0.2010,  0.6292, -0.8918,  0.3371
-0.7390,  0.6849,  0.2367,  0.0626, -0.5034,  0.7039
-0.1567, -0.8711,  0.7940, -0.5932,  0.6525,  0.1710
 0.7635, -0.0265,  0.1969,  0.0545,  0.2496,  0.1445
 0.7675,  0.1354, -0.7698, -0.5460,  0.1920,  0.1728
-0.5211, -0.7372, -0.6763,  0.6897,  0.2044,  0.5217
 0.1913,  0.1980,  0.2314, -0.8816,  0.5006,  0.1998
 0.8964,  0.0694, -0.6149,  0.5059, -0.9854,  0.1825
 0.1767,  0.7104,  0.2093,  0.6452,  0.7590,  0.2832
-0.3580, -0.7541,  0.4426, -0.1193, -0.7465,  0.5657
-0.5996,  0.5766, -0.9758, -0.3933, -0.9572,  0.6800
 0.9950,  0.1641, -0.4132,  0.8579,  0.0142,  0.2003
-0.4717, -0.3894, -0.2567, -0.5111,  0.1691,  0.4266
 0.3917, -0.8561,  0.9422,  0.5061,  0.6123,  0.1212
-0.0366, -0.1087,  0.3449, -0.1025,  0.4086,  0.2475
 0.3633,  0.3943,  0.2372, -0.6980,  0.5216,  0.1925
-0.5325, -0.6466, -0.2178, -0.3589,  0.6310,  0.3568
 0.2271,  0.5200, -0.1447, -0.8011, -0.7699,  0.3128
 0.6415,  0.1993,  0.3777, -0.0178, -0.8237,  0.2181
-0.5298, -0.0768, -0.6028, -0.9490,  0.4588,  0.4356
 0.6870, -0.1431,  0.7294,  0.3141,  0.1621,  0.1632
-0.5985,  0.0591,  0.7889, -0.3900,  0.7419,  0.2945
 0.3661,  0.7984, -0.8486,  0.7572, -0.6183,  0.3449
 0.6995,  0.3342, -0.3113, -0.6972,  0.2707,  0.1712
 0.2565,  0.9126,  0.1798, -0.6043, -0.1413,  0.2893
-0.3265,  0.9839, -0.2395,  0.9854,  0.0376,  0.4770
 0.2690, -0.1722,  0.9818,  0.8599, -0.7015,  0.3954
-0.2102, -0.0768,  0.1219,  0.5607, -0.0256,  0.3949
 0.8216, -0.9555,  0.6422, -0.6231,  0.3715,  0.0801
-0.2896,  0.9484, -0.7545, -0.6249,  0.7789,  0.4370
-0.9985, -0.5448, -0.7092, -0.5931,  0.7926,  0.5402

Test data:

# synthetic_test_40.txt
#
 0.7462,  0.4006, -0.0590,  0.6543, -0.0083,  0.1935
 0.8495, -0.2260, -0.0142, -0.4911,  0.7699,  0.1078
-0.2335, -0.4049,  0.4352, -0.6183, -0.7636,  0.5088
 0.1810, -0.5142,  0.2465,  0.2767, -0.3449,  0.3136
-0.8650,  0.7611, -0.0801,  0.5277, -0.4922,  0.7140
-0.2358, -0.7466, -0.5115, -0.8413, -0.3943,  0.4533
 0.4834,  0.2300,  0.3448, -0.9832,  0.3568,  0.1360
-0.6502, -0.6300,  0.6885,  0.9652,  0.8275,  0.3046
-0.3053,  0.5604,  0.0929,  0.6329, -0.0325,  0.4756
-0.7995,  0.0740, -0.2680,  0.2086,  0.9176,  0.4565
-0.2144, -0.2141,  0.5813,  0.2902, -0.2122,  0.4119
-0.7278, -0.0987, -0.3312, -0.5641,  0.8515,  0.4438
 0.3793,  0.1976,  0.4933,  0.0839,  0.4011,  0.1905
-0.8568,  0.9573, -0.5272,  0.3212, -0.8207,  0.7415
-0.5785,  0.0056, -0.7901, -0.2223,  0.0760,  0.5551
 0.0735, -0.2188,  0.3925,  0.3570,  0.3746,  0.2191
 0.1230, -0.2838,  0.2262,  0.8715,  0.1938,  0.2878
 0.4792, -0.9248,  0.5295,  0.0366, -0.9894,  0.3149
-0.4456,  0.0697,  0.5359, -0.8938,  0.0981,  0.3879
 0.8629, -0.8505, -0.4464,  0.8385,  0.5300,  0.1769
 0.1995,  0.6659,  0.7921,  0.9454,  0.9970,  0.2330
-0.0249, -0.3066, -0.2927, -0.4923,  0.8220,  0.2437
 0.4513, -0.9481, -0.0770, -0.4374, -0.9421,  0.2879
-0.3405,  0.5931, -0.3507, -0.3842,  0.8562,  0.3987
 0.9538,  0.0471,  0.9039,  0.7760,  0.0361,  0.1706
-0.0887,  0.2104,  0.9808,  0.5478, -0.3314,  0.4128
-0.8220, -0.6302,  0.0537, -0.1658,  0.6013,  0.4306
-0.4123, -0.2880,  0.9074, -0.0461, -0.4435,  0.5144
 0.0060,  0.2867, -0.7775,  0.5161,  0.7039,  0.3599
-0.7968, -0.5484,  0.9426, -0.4308,  0.8148,  0.2979
 0.7811,  0.8450, -0.6877,  0.7594,  0.2640,  0.2362
-0.6802, -0.1113, -0.8325, -0.6694, -0.6056,  0.6544
 0.3821,  0.1476,  0.7466, -0.5107,  0.2592,  0.1648
 0.7265,  0.9683, -0.9803, -0.4943, -0.5523,  0.2454
-0.9049, -0.9797, -0.0196, -0.9090, -0.4433,  0.6447
-0.4607,  0.1811, -0.2389,  0.4050, -0.0078,  0.5229
 0.2664, -0.2932, -0.4259, -0.7336,  0.8742,  0.1834
-0.4507,  0.1029, -0.6294, -0.1158, -0.6294,  0.6081
 0.8948, -0.0124,  0.9278,  0.2899, -0.0314,  0.1534
-0.1323, -0.8813, -0.0146, -0.0697,  0.6135,  0.2386
Posted in Machine Learning | Leave a comment

Example of Anomaly Detection Using scikit IsolationForest

Posted on June 19, 2026 by jamesdmccaffrey

I came across a relatively obscure module in the scikit-learn library called IsolationForest. The module is used for anomaly detection. In a nutshell, if you apply a decision tree to a dataset, because of the way the branches are created, after the tree is created, tree nodes that are near the root, and which hold a single value, are anomalous.

I decided to put a demo together.

For my demo I used a set of synthetic data that looks like:

-0.1660,  0.4406, -0.9998, -0.3953, -0.7065,  0.4840
 0.9776, -0.9616,  0.9704, -0.9911,  0.9562,  0.1568
-0.9452,  0.3409, -0.1654,  0.1174, -0.7192,  0.8054
 0.9365, -0.3732,  0.3846,  0.7528,  0.7892,  0.1345
-0.8299, -0.9219, -0.6603,  0.7563, -0.8033,  0.7955
. . .

The last value in each column is a dependent variable, used for regression explorations, so I used only the first five values on each line. There are 200 data items.

I deliberately modified data item [1] to make the magnitudes of each value very large — so data item [1] is anomalous. Notice all the values in item [1] are very close to -1 or +1.

The output of my IsolationForest demo is:

Begin scikit IsolationForest demo

Loading synthetic data (200)

First three data items:
[-0.1660  0.4406 -0.9998 -0.3953 -0.7065]
[ 0.9776 -0.9616  0.9704 -0.9911  0.9562]
[-0.9452  0.3409 -0.1654  0.1174 -0.7192]

Creating IsolationForest
Done

Analyzing dataset
Done

First three anomaly scores
  0.0044
 -0.1226
  0.0079

Most anomalous item = [1]
Anomaly score = -0.1226

End demo

Smaller values (more negative) are more anomalous. The demo correctly identified item [1] as the most anomalous.

When I first discovered the scikit IsolationForest, I didn’t like some of the technical details and so I used the key ideas to implement a C# system that I called an Anomaly Forest. When I ran the data through the C# program, it also picked data item [1] as most anomalous. (See the screenshot above).

Different anomaly detection algorithms identify different types of anomalies. The scikit IsolationForest module is just one of many techniques.


In some sense, machine learning is all about finding hidden patterns in data. Every cover of Playboy Magazine (except for the very first issue in December 1953) has a logo of a bunny. On most covers, the bunny logo is fairly obvious, but sometimes the logo is very cleverly hidden.

Left: On the July 1976 (the 200th anniversary of the United States), the bunny logo is disguised as one of the stars on the U.S. flag.

Right: This is the most famous cover for the hidden logo. Many thousands of readers wrote to the magazine saying that they had examined every square inch of the cover and there is no logo. But there is — the trickiest hidden logo ever, up until that time. The logo is hidden in the text “The Playmate of the Year”. The bunny head is the upper-left part of the “Y” of “Year” and the ears are the bottom part of the “P” and the “y” in Playmate”. Because of this tricky cover, starting a few months later, the magazine explained where the hidden logo is on the Contents page of each issue.

Demo program.

# isolation_forest_scikit.py
# anomaly detection

import numpy as np
from sklearn.ensemble import IsolationForest

np.set_printoptions(precision=4, suppress=True,
  floatmode='fixed', linewidth=120)

print("\nBegin scikit IsolationForest demo ")

print("\nLoading synthetic data (200) ")
data_file = ".\\Data\\synthetic_200.txt"

cols_X = [0,1,2,3,4]  
data_X = np.loadtxt(data_file, comments="#",
  usecols=cols_X, delimiter=",",  dtype=np.float64)

print("\nFirst three data items: ")
for i in range(3):
  print(data_X[i])

print("\nCreating IsolationForest ")
detector = IsolationForest(n_estimators=50, random_state=0)
print("Done ")

print("\nAnalyzing dataset ")
detector.fit(data_X)
print("Done ")

scores = detector.decision_function(data_X)
print("\nFirst three anomaly scores ")

for i in range(3):
  print("%8.4f " % scores[i])

min_indx = 0
min_score = scores[0]
for i in range(len(scores)):
  if scores[i] < min_score:
    min_score = scores[i]
    min_indx = i

print("\nMost anomalous item = [" + str(min_indx) + "]")
print("Anomaly score = %0.4f " % min_score)

# print("")
# for i in range(len(scores)):
#   print("%4d  %0.4f " % (i, scores[i]))

print("\nEnd demo ")

Data:

# synthetic_200.txt
#
-0.1660,  0.4406, -0.9998, -0.3953, -0.7065,  0.4840
 0.9776, -0.9616,  0.9704, -0.9911,  0.9562,  0.1568
-0.9452,  0.3409, -0.1654,  0.1174, -0.7192,  0.8054
 0.9365, -0.3732,  0.3846,  0.7528,  0.7892,  0.1345
-0.8299, -0.9219, -0.6603,  0.7563, -0.8033,  0.7955
 0.0663,  0.3838, -0.3690,  0.3730,  0.6693,  0.3206
-0.9634,  0.5003,  0.9777,  0.4963, -0.4391,  0.7377
-0.1042,  0.8172, -0.4128, -0.4244, -0.7399,  0.4801
-0.9613,  0.3577, -0.5767, -0.4689, -0.0169,  0.6861
-0.7065,  0.1786,  0.3995, -0.7953, -0.1719,  0.5569
 0.3888, -0.1716, -0.9001,  0.0718,  0.3276,  0.2500
 0.1731,  0.8068, -0.7251, -0.7214,  0.6148,  0.3297
-0.2046, -0.6693,  0.8550, -0.3045,  0.5016,  0.2129
 0.2473,  0.5019, -0.3022, -0.4601,  0.7918,  0.2613
-0.1438,  0.9297,  0.3269,  0.2434, -0.7705,  0.5171
 0.1568, -0.1837, -0.5259,  0.8068,  0.1474,  0.3307
-0.9943,  0.2343, -0.3467,  0.0541,  0.7719,  0.5581
 0.2467, -0.9684,  0.8589,  0.3818,  0.9946,  0.1092
-0.6553, -0.7257,  0.8652,  0.3936, -0.8680,  0.7018
 0.8460,  0.4230, -0.7515, -0.9602, -0.9476,  0.1996
-0.9434, -0.5076,  0.7201,  0.0777,  0.1056,  0.5664
 0.9392,  0.1221, -0.9627,  0.6013, -0.5341,  0.1533
 0.6142, -0.2243,  0.7271,  0.4942,  0.1125,  0.1661
 0.4260,  0.1194, -0.9749, -0.8561,  0.9346,  0.2230
 0.1362, -0.5934, -0.4953,  0.4877, -0.6091,  0.3810
 0.6937, -0.5203, -0.0125,  0.2399,  0.6580,  0.1460
-0.6864, -0.9628, -0.8600, -0.0273,  0.2127,  0.5387
 0.9772,  0.1595, -0.2397,  0.1019,  0.4907,  0.1611
 0.3385, -0.4702, -0.8673, -0.2598,  0.2594,  0.2270
-0.8669, -0.4794,  0.6095, -0.6131,  0.2789,  0.4700
 0.0493,  0.8496, -0.4734, -0.8681,  0.4701,  0.3516
 0.8639, -0.9721, -0.5313,  0.2336,  0.8980,  0.1412
 0.9004,  0.1133,  0.8312,  0.2831, -0.2200,  0.1782
 0.0991,  0.8524,  0.8375, -0.2102,  0.9265,  0.2150
-0.6521, -0.7473, -0.7298,  0.0113, -0.9570,  0.7422
 0.6190, -0.3105,  0.8802,  0.1640,  0.7577,  0.1056
 0.6895,  0.8108, -0.0802,  0.0927,  0.5972,  0.2214
 0.1982, -0.9689,  0.1870, -0.1326,  0.6147,  0.1310
-0.3695,  0.7858,  0.1557, -0.6320,  0.5759,  0.3773
-0.1596,  0.3581,  0.8372, -0.9992,  0.9535,  0.2071
-0.2468,  0.9476,  0.2094,  0.6577,  0.1494,  0.4132
 0.1737,  0.5000,  0.7166,  0.5102,  0.3961,  0.2611
 0.7290, -0.3546,  0.3416, -0.0983, -0.2358,  0.1332
-0.3652,  0.2438, -0.1395,  0.9476,  0.3556,  0.4170
-0.6029, -0.1466, -0.3133,  0.5953,  0.7600,  0.4334
-0.4596, -0.4953,  0.7098,  0.0554,  0.6043,  0.2775
 0.1450,  0.4663,  0.0380,  0.5418,  0.1377,  0.2931
-0.8636, -0.2442, -0.8407,  0.9656, -0.6368,  0.7429
 0.6237,  0.7499,  0.3768,  0.1390, -0.6781,  0.2185
-0.5499,  0.1850, -0.3755,  0.8326,  0.8193,  0.4399
-0.4858, -0.7782, -0.6141, -0.0008,  0.4572,  0.4197
 0.7033, -0.1683,  0.2334, -0.5327, -0.7961,  0.1776
 0.0317, -0.0457, -0.6947,  0.2436,  0.0880,  0.3345
 0.5031, -0.5559,  0.0387,  0.5706, -0.9553,  0.3107
-0.3513,  0.7458,  0.6894,  0.0769,  0.7332,  0.3170
 0.2205,  0.5992, -0.9309,  0.5405,  0.4635,  0.3532
-0.4806, -0.4859,  0.2646, -0.3094,  0.5932,  0.3202
 0.9809, -0.3995, -0.7140,  0.8026,  0.0831,  0.1600
 0.9495,  0.2732,  0.9878,  0.0921,  0.0529,  0.1289
-0.9476, -0.6792,  0.4913, -0.9392, -0.2669,  0.5966
 0.7247,  0.3854,  0.3819, -0.6227, -0.1162,  0.1550
-0.5922, -0.5045, -0.4757,  0.5003, -0.0860,  0.5863
-0.8861,  0.0170, -0.5761,  0.5972, -0.4053,  0.7301
 0.6877, -0.2380,  0.4997,  0.0223,  0.0819,  0.1404
 0.9189,  0.6079, -0.9354,  0.4188, -0.0700,  0.1907
-0.1428, -0.7820,  0.2676,  0.6059,  0.3936,  0.2790
 0.5324, -0.3151,  0.6917, -0.1425,  0.6480,  0.1071
-0.8432, -0.9633, -0.8666, -0.0828, -0.7733,  0.7784
-0.9444,  0.5097, -0.2103,  0.4939, -0.0952,  0.6787
-0.0520,  0.6063, -0.1952,  0.8094, -0.9259,  0.4836
 0.5477, -0.7487,  0.2370, -0.9793,  0.0773,  0.1241
 0.2450,  0.8116,  0.9799,  0.4222,  0.4636,  0.2355
 0.8186, -0.1983, -0.5003, -0.6531, -0.7611,  0.1511
-0.4714,  0.6382, -0.3788,  0.9648, -0.4667,  0.5950
 0.0673, -0.3711,  0.8215, -0.2669, -0.1328,  0.2677
-0.9381,  0.4338,  0.7820, -0.9454,  0.0441,  0.5518
-0.3480,  0.7190,  0.1170,  0.3805, -0.0943,  0.4724
-0.9813,  0.1535, -0.3771,  0.0345,  0.8328,  0.5438
-0.1471, -0.5052, -0.2574,  0.8637,  0.8737,  0.3042
-0.5454, -0.3712, -0.6505,  0.2142, -0.1728,  0.5783
 0.6327, -0.6297,  0.4038, -0.5193,  0.1484,  0.1153
-0.5424,  0.3282, -0.0055,  0.0380, -0.6506,  0.6613
 0.1414,  0.9935,  0.6337,  0.1887,  0.9520,  0.2540
-0.9351, -0.8128, -0.8693, -0.0965, -0.2491,  0.7353
 0.9507, -0.6640,  0.9456,  0.5349,  0.6485,  0.1059
-0.0462, -0.9737, -0.2940, -0.0159,  0.4602,  0.2606
-0.0627, -0.0852, -0.7247, -0.9782,  0.5166,  0.2977
 0.0478,  0.5098, -0.0723, -0.7504, -0.3750,  0.3335
 0.0090,  0.3477,  0.5403, -0.7393, -0.9542,  0.4415
-0.9748,  0.3449,  0.3736, -0.1015,  0.8296,  0.4358
 0.2887, -0.9895, -0.0311,  0.7186,  0.6608,  0.2057
 0.1570, -0.4518,  0.1211,  0.3435, -0.2951,  0.3244
 0.7117, -0.6099,  0.4946, -0.4208,  0.5476,  0.1096
-0.2929, -0.5726,  0.5346, -0.3827,  0.4665,  0.2465
 0.4889, -0.5572, -0.5718, -0.6021, -0.7150,  0.2163
-0.7782,  0.3491,  0.5996, -0.8389, -0.5366,  0.6516
-0.5847,  0.8347,  0.4226,  0.1078, -0.3910,  0.6134
 0.8469,  0.4121, -0.0439, -0.7476,  0.9521,  0.1571
-0.6803, -0.5948, -0.1376, -0.1916, -0.7065,  0.7156
 0.2878,  0.5086, -0.5785,  0.2019,  0.4979,  0.2980
 0.2764,  0.1943, -0.4090,  0.4632,  0.8906,  0.2960
-0.8877,  0.6705, -0.6155, -0.2098, -0.3998,  0.7107
-0.8398,  0.8093, -0.2597,  0.0614, -0.0118,  0.6502
-0.8476,  0.0158, -0.4769, -0.2859, -0.7839,  0.7715
 0.5751, -0.7868,  0.9714, -0.6457,  0.1448,  0.1175
 0.4802, -0.7001,  0.1022, -0.5668,  0.5184,  0.1090
 0.4458, -0.6469,  0.7239, -0.9604,  0.7205,  0.0779
 0.5175,  0.4339,  0.9747, -0.4438, -0.9924,  0.2879
 0.8678,  0.7158,  0.4577,  0.0334,  0.4139,  0.1678
 0.5406,  0.5012,  0.2264, -0.1963,  0.3946,  0.2088
-0.9938,  0.5498,  0.7928, -0.5214, -0.7585,  0.7687
 0.7661,  0.0863, -0.4266, -0.7233, -0.4197,  0.1466
 0.2277, -0.3517, -0.0853, -0.1118,  0.6563,  0.1767
 0.3499, -0.5570, -0.0655, -0.3705,  0.2537,  0.1632
 0.7547, -0.1046,  0.5689, -0.0861,  0.3125,  0.1257
 0.8186,  0.2110,  0.5335,  0.0094, -0.0039,  0.1391
 0.6858, -0.8644,  0.1465,  0.8855,  0.0357,  0.1845
-0.4967,  0.4015,  0.0805,  0.8977,  0.2487,  0.4663
 0.6760, -0.9841,  0.9787, -0.8446, -0.3557,  0.1509
-0.1203, -0.4885,  0.6054, -0.0443, -0.7313,  0.4854
 0.8557,  0.7919, -0.0169,  0.7134, -0.1628,  0.2002
 0.0115, -0.6209,  0.9300, -0.4116, -0.7931,  0.4052
-0.7114, -0.9718,  0.4319,  0.1290,  0.5892,  0.3661
 0.3915,  0.5557, -0.1870,  0.2955, -0.6404,  0.2954
-0.3564, -0.6548, -0.1827, -0.5172, -0.1862,  0.4622
 0.2392, -0.4959,  0.5857, -0.1341, -0.2850,  0.2470
-0.3394,  0.3947, -0.4627,  0.6166, -0.4094,  0.5325
 0.7107,  0.7768, -0.6312,  0.1707,  0.7964,  0.2757
-0.1078,  0.8437, -0.4420,  0.2177,  0.3649,  0.4028
-0.3139,  0.5595, -0.6505, -0.3161, -0.7108,  0.5546
 0.4335,  0.3986,  0.3770, -0.4932,  0.3847,  0.1810
-0.2562, -0.2894, -0.8847,  0.2633,  0.4146,  0.4036
 0.2272,  0.2966, -0.6601, -0.7011,  0.0284,  0.2778
-0.0743, -0.1421, -0.0054, -0.6770, -0.3151,  0.3597
-0.4762,  0.6891,  0.6007, -0.1467,  0.2140,  0.4266
-0.4061,  0.7193,  0.3432,  0.2669, -0.7505,  0.6147
-0.0588,  0.9731,  0.8966,  0.2902, -0.6966,  0.4955
-0.0627, -0.1439,  0.1985,  0.6999,  0.5022,  0.3077
 0.1587,  0.8494, -0.8705,  0.9827, -0.8940,  0.4263
-0.7850,  0.2473, -0.9040, -0.4308, -0.8779,  0.7199
 0.4070,  0.3369, -0.2428, -0.6236,  0.4940,  0.2215
-0.0242,  0.0513, -0.9430,  0.2885, -0.2987,  0.3947
-0.5416, -0.1322, -0.2351, -0.0604,  0.9590,  0.3683
 0.1055,  0.7783, -0.2901, -0.5090,  0.8220,  0.2984
-0.9129,  0.9015,  0.1128, -0.2473,  0.9901,  0.4776
-0.9378,  0.1424, -0.6391,  0.2619,  0.9618,  0.5368
 0.7498, -0.0963,  0.4169,  0.5549, -0.0103,  0.1614
-0.2612, -0.7156,  0.4538, -0.0460, -0.1022,  0.3717
 0.7720,  0.0552, -0.1818, -0.4622, -0.8560,  0.1685
-0.4177,  0.0070,  0.9319, -0.7812,  0.3461,  0.3052
-0.0001,  0.5542, -0.7128, -0.8336, -0.2016,  0.3803
 0.5356, -0.4194, -0.5662, -0.9666, -0.2027,  0.1776
-0.2378,  0.3187, -0.8582, -0.6948, -0.9668,  0.5474
-0.1947, -0.3579,  0.1158,  0.9869,  0.6690,  0.2992
 0.3992,  0.8365, -0.9205, -0.8593, -0.0520,  0.3154
-0.0209,  0.0793,  0.7905, -0.1067,  0.7541,  0.1864
-0.4928, -0.4524, -0.3433,  0.0951, -0.5597,  0.6261
-0.8118,  0.7404, -0.5263, -0.2280,  0.1431,  0.6349
 0.0516, -0.8480,  0.7483,  0.9023,  0.6250,  0.1959
-0.3212,  0.1093,  0.9488, -0.3766,  0.3376,  0.2735
-0.3481,  0.5490, -0.3484,  0.7797,  0.5034,  0.4379
-0.5785, -0.9170, -0.3563, -0.9258,  0.3877,  0.4121
 0.3407, -0.1391,  0.5356,  0.0720, -0.9203,  0.3458
-0.3287, -0.8954,  0.2102,  0.0241,  0.2349,  0.3247
-0.1353,  0.6954, -0.0919, -0.9692,  0.7461,  0.3338
 0.9036, -0.8982, -0.5299, -0.8733, -0.1567,  0.1187
 0.7277, -0.8368, -0.0538, -0.7489,  0.5458,  0.0830
 0.9049,  0.8878,  0.2279,  0.9470, -0.3103,  0.2194
 0.7957, -0.1308, -0.5284,  0.8817,  0.3684,  0.2172
 0.4647, -0.4931,  0.2010,  0.6292, -0.8918,  0.3371
-0.7390,  0.6849,  0.2367,  0.0626, -0.5034,  0.7039
-0.1567, -0.8711,  0.7940, -0.5932,  0.6525,  0.1710
 0.7635, -0.0265,  0.1969,  0.0545,  0.2496,  0.1445
 0.7675,  0.1354, -0.7698, -0.5460,  0.1920,  0.1728
-0.5211, -0.7372, -0.6763,  0.6897,  0.2044,  0.5217
 0.1913,  0.1980,  0.2314, -0.8816,  0.5006,  0.1998
 0.8964,  0.0694, -0.6149,  0.5059, -0.9854,  0.1825
 0.1767,  0.7104,  0.2093,  0.6452,  0.7590,  0.2832
-0.3580, -0.7541,  0.4426, -0.1193, -0.7465,  0.5657
-0.5996,  0.5766, -0.9758, -0.3933, -0.9572,  0.6800
 0.9950,  0.1641, -0.4132,  0.8579,  0.0142,  0.2003
-0.4717, -0.3894, -0.2567, -0.5111,  0.1691,  0.4266
 0.3917, -0.8561,  0.9422,  0.5061,  0.6123,  0.1212
-0.0366, -0.1087,  0.3449, -0.1025,  0.4086,  0.2475
 0.3633,  0.3943,  0.2372, -0.6980,  0.5216,  0.1925
-0.5325, -0.6466, -0.2178, -0.3589,  0.6310,  0.3568
 0.2271,  0.5200, -0.1447, -0.8011, -0.7699,  0.3128
 0.6415,  0.1993,  0.3777, -0.0178, -0.8237,  0.2181
-0.5298, -0.0768, -0.6028, -0.9490,  0.4588,  0.4356
 0.6870, -0.1431,  0.7294,  0.3141,  0.1621,  0.1632
-0.5985,  0.0591,  0.7889, -0.3900,  0.7419,  0.2945
 0.3661,  0.7984, -0.8486,  0.7572, -0.6183,  0.3449
 0.6995,  0.3342, -0.3113, -0.6972,  0.2707,  0.1712
 0.2565,  0.9126,  0.1798, -0.6043, -0.1413,  0.2893
-0.3265,  0.9839, -0.2395,  0.9854,  0.0376,  0.4770
 0.2690, -0.1722,  0.9818,  0.8599, -0.7015,  0.3954
-0.2102, -0.0768,  0.1219,  0.5607, -0.0256,  0.3949
 0.8216, -0.9555,  0.6422, -0.6231,  0.3715,  0.0801
-0.2896,  0.9484, -0.7545, -0.6249,  0.7789,  0.4370
-0.9985, -0.5448, -0.7092, -0.5931,  0.7926,  0.5402
Posted in Machine Learning, Scikit | Leave a comment

“The AI Coding Revolution: Productivity Boom and Employment Crisis” on the Pure AI Web Site

Posted on June 18, 2026 by jamesdmccaffrey

I contributed some technical content and opinions to an article titled “The AI Coding Revolution: Productivity Boom and Employment Crisis” on the Pure AI web site. See https://pureai.com/articles/2026/06/01/the-ai-coding-revolution-productivity-boom-and-employment-crisis.aspx.

AI-assisted coding is easily the biggest technology change that I’ve seen in my 50+ years of writing code. I helped write an article that discusses what some of the impacts are. Briefly:

1.) AI coding assistants are rapidly evolving from code-completion tools into autonomous software development agents.

2.) The technology is increasing software development productivity while reducing demand for routine programming work.

3.) The long-term benefits of AI-generated software may be significant, but the short-term impact on technology employment will be severe.

McCaffrey commented, “AI-assisted coding is particularly effective in SQL development and web development, where large amounts of structured training data and established coding patterns exist. AI coding is weakest at user interface design, which is highly subjective and a moving target.”

The AI coding market is currently dominated by six powerful competitors: Microsoft, Google, Anthropic, Amazon, OpenAI, and Cursor.

McCaffrey said, “Microsoft’s GitHub Copilot is the dominant enterprise solution. Google’s Gemini Code Assist is known for its ability to analyze extremely large codebases. Anthropic’s Claude Code has emerged as a leading agentic coding platform capable of operating autonomously.”

“Amazon Q Developer specializes in cloud-native development within AWS environments. OpenAI has evolved from primarily supplying the underlying models for other products to offering its own autonomous coding agents through Codex.”

“Cursor has become one of the fastest-growing software products in history, transforming from an AI-powered editor into a platform where developers manage teams of autonomous software agents, using large language models from different companies.”

Industry layoff trackers estimate that well over 400,000 technology workers were laid off in the U.S. during the 18 months between January 2025 and June 2026.

McCaffrey commented, “AI coding assistants are fueling a technology employment Armageddon. Some research firms suggest that AI will eventually create more tech jobs than it eliminates. I am extremely pessimistic about that claim, at least for the short term of the next five years or so.”

“The hundreds of thousands of jobs that have already been eliminated will not reappear in another form for the foreseeable future, and many tens or even hundreds of thousands more tech jobs will be wiped out in the near future.”


AI coding is having a profound effect on society.

The early James Bond films had a profound effect on culture. The early movies were released in the 1960s, just a few years after the end of World War II, at a time when Japan was almost completely unknown to audiences. These early films have not held up too well over time, but their impact at the time was huge.

Left: In “Thunderball” (1965), Bond briefly cooperates with Madame LaPorte (actress Maryse Guy Mitsouko), a French secret service agent. Bond must recover two stolen nuclear bombs that have been hidden in the Caribbean. He does.

Right: In “You Only Live Twice” (1967), Bond cooperates with Aki (actress Akiko Wakabayashi), a Japanese secret service agent. Bond must stop a plan by the evil SPECTRE organization to start a nuclear war between the U.S. and the Soviet Union. He does.

Posted in Uncategorized | Leave a comment

Why I Never Use a Voting Regressor Model for Machine Learning Regression

Posted on June 17, 2026 by jamesdmccaffrey

Bottom line: I never use a voting regressor model because the number of meta parameters (which regression models to use) and model parameters (such as a model learning rate) is astronomically large and therefore unmanageable in practice.

A voting regressor model is conceptually simple. Instead of using a particular model (linear regression, quadratic regression, kernel ridge, nearest neighbors, neural network, random forest, adaptive boosting, gradient boosting, etc., etc., etc.), you use several models and then use the average of their predictions.

The problem is that there are just too many parameters. First you have to choose which models to use, and then each model has on average about three to five parameters to specify. This just isn’t feasible.

I put together a demo using the VotingRegressor model from the scikit-learn library. For my demo I used the Diabetes Dataset, and I used three models: nearest neighbors, kernel ridge, gradient boosting. I knew from previous explorations that the default target in the Diabetes Dataset cannot be predicted with meaningful accuracy. As expected, the VotingRegressor model failed to give good accuracy.

The raw Diabetes Dataset looks like:

59, 2, 32.1, 101.00, 157,  93.2, 38, 4.00, 4.8598, 87, 151
48, 1, 21.6,  87.00, 183, 103.2, 70, 3.00, 3.8918, 69,  75
72, 2, 30.5,  93.00, 156,  93.6, 41, 4.00, 4.6728, 85, 141
. . .

The dataset has 442 items. Each item represents a patient and has 10 predictor values followed by a target value to predict. The 10 predictor variables are age in column [0], sex [1], body mass index [2], blood pressure [3], serum cholesterol [4], low-density lipoproteins [5], high-density lipoproteins [6], total cholesterol [7], triglycerides [8], blood sugar [9]. The stated (in the documentation) target value to predict in the last column is a measure of diabetes [10].

Note: The sex encoding isn’t explained but I suspect male = 1, female = 2 because there are 235 1 values and 206 2 values). Also, the scikit Diabetes Dataset is not the same as the Pima Diabetes dataset.

I re-encoded the sex predictor from 1,2 to 0,1. I normalized the other predictor values using divide-by-constant normalization (100, 1, 100, 1000, 1000, 1000, 100, 10, 10, 1000, 1000). I normalized the target y values by dividing by 1000. The resulting data is:

0.5900, 1.0000, 0.3210, . . . 0.1510
0.4800, 0.0000, 0.2160, . . . 0.0750
0.7200, 1.0000, 0.3050, . . . 0.1410
. . .

I used the first 342 data items for training and the remaining 100 items for testing. I experimented with other splitting ideas and normalization techniques, but none had any significant effect.

The output of my demo is:

Begin Diabetes Dataset using VotingRegressor

Loading diabetes train (342), test (100) data
Done

First three X predictors:
[0.5900 1.0000 0.3210 0.1010 0.1570 . . . 0.0870]
[0.4800 0.0000 0.2160 0.0870 0.1830 . . . 0.0690]
[0.7200 1.0000 0.3050 0.0930 0.1560 . . . 0.0850]

First three y targets:
0.1510
0.0750
0.1410

Creating nearest neighbors, kernel ridge, 
  gradient boost regressors

Creating voting regressor model
Done

acc train (0.10) = 0.2895
acc test (0.10) = 0.1630

End demo

The voting regressor model had poor accuracy as expected. A prediction is scored as correct if it’s within 10% of the true target value.

If you use the built-in load_diabetes() function with a return_X_y=True parameter, column [10] is automatically the target to predict. But I discovered that columns [4], [5], [6], [7], and [8] can be predicted meaningfully. You can load the data as a DataFrame and then specify a different column as the target. Alternatively, you can use preprocessed training data.

The key lines in the demo program are:

. . .
r1 = KNeighborsRegressor(n_neighbors=3)
r2 = KernelRidge(kernel='rbf', gamma=0.005, alpha=0.01)
r3 = GradientBoostingRegressor(n_estimators=100,
  max_depth=5, learning_rate=0.10, random_state=0)

model = VotingRegressor([('knn', r1), ('krr', r2), ('gbr', r3)])
model.fit(train_X, train_y)
. . .

An interesting exploration.


Different regression techniques have different characteristics, and unfortunately, there’s no principled way to select which techniques to use in voting ression.

My two favorite movie genres are science fiction, and comedy. Here are three films that, in my opinion, have a good balance of science fiction and comedy. I’ve noticed that opinion among my friends about these three movies is quite polarized.

Left: In “Galaxy Quest” (1999), a group of past-their-prime actors from a Star Trek like TV series are recruited by aliens to save them from a really evil other alien race. I give this movie my personal A grade.

Center: In “The Fifth Element” (1997), a retired soldier is called back to service to literally save the Earth. He is aided by a female alien. I give this movie my personal A grade. I actually have a tattoo of the four elemental symbols used in the movie on my back. Seriously, I really do.

Right: In “Dude, Where’s My Car?” (2000), two losers have all kinds of outlandish adventures and meet different alien races, while saving the Universe from extinction. I give this movie my personal B+ grade.

Demo program. Replace the “lt” in the accuracy() function with the Boolean less-than symbol (my blog editor chokes on symbols).

# diabetes_voting_scikit.py
# various techniques for the Diabetes Dataset

import numpy as np
from sklearn.neighbors import KNeighborsRegressor
from sklearn.kernel_ridge import KernelRidge
# from sklearn.tree import DecisionTreeRegressor
from sklearn.ensemble import GradientBoostingRegressor
from sklearn.ensemble import VotingRegressor

np.set_printoptions(precision=4, suppress=True,
  floatmode='fixed', linewidth=120)

# -----------------------------------------------------------

def accuracy(model, data_X, data_y, pct_close):
  n = len(data_X)
  n_correct = 0; n_wrong = 0
  for i in range(n):
    x = data_X[i].reshape(1,-1)
    y = data_y[i]
    y_pred = model.predict(x)

    if np.abs(y - y_pred) "lt" np.abs(y * pct_close):
      n_correct += 1
    else: 
      n_wrong += 1
  # print("Correct = " + str(n_correct))
  # print("Wrong   = " + str(n_wrong))
  return n_correct / (n_correct + n_wrong)

# -----------------------------------------------------------

def MSE(model, data_X, data_y):
  n = len(data_X)
  sum = 0.0
  for i in range(n):
    x = data_X[i].reshape(1,-1)
    y = data_y[i]
    y_pred = model.predict(x)[0]
    # print(y_pred); input()
    sum += (y - y_pred) * (y - y_pred)

  return sum / n

# -----------------------------------------------------------

print("\nBegin Diabetes Dataset using VotingRegressor ")

print("\nLoading diabetes train (342), test (100) data ")
train_file = ".\\Data\\diabetes_norm_train_342.txt"

cols_X = [0,1,2,3,4,5,6,7,8,9]  
col_y = 10  # cols # 4 5 6 7 8 are much better
train_X = np.loadtxt(train_file, comments="#",
  usecols=cols_X, delimiter=",",  dtype=np.float64)
train_y = np.loadtxt(train_file, comments="#",
  usecols=col_y, delimiter=",",  dtype=np.float64)

test_file = ".\\Data\\diabetes_norm_test_100.txt"
test_X = np.loadtxt(test_file, comments="#",
  usecols=cols_X, delimiter=",",  dtype=np.float64)
test_y = np.loadtxt(test_file, comments="#",
  usecols=col_y, delimiter=",",  dtype=np.float64)
print("Done ")

# alternative normalization and split
# from sklearn.datasets import load_diabetes
# from sklearn.model_selection import train_test_split
# X, y = load_diabetes(return_X_y=True, scaled=True)
# train_X, test_X, train_y, test_y = \
#   train_test_split(X, y, random_state=0)  # 25% test

print("\nFirst three X predictors: ")
for i in range(3):
  print(train_X[i])
print("\nFirst three y targets: ")
for i in range(3):
  print("%0.4f" % train_y[i])

print("\nCreating nearest neighbors, kernel ridge, " + \
  "gradient boost regressors ")
r1 = KNeighborsRegressor(n_neighbors=3)
r2 = KernelRidge(kernel='rbf', gamma=0.005, alpha=0.01)
r3 = GradientBoostingRegressor(n_estimators=100,
  max_depth=5, learning_rate=0.10, random_state=0)

print("\nCreating voting regressor model ")
model = VotingRegressor([('knn', r1), ('krr', r2), ('gbr', r3)])
model.fit(train_X, train_y)
print("Done ")

acc_train = accuracy(model, train_X, train_y, 0.10)
print("\nacc train (0.10) = %0.4f " % acc_train)
acc_test = accuracy(model, test_X, test_y, 0.10)
print("acc test (0.10) = %0.4f " % acc_test)

print("\nEnd demo ")

Training data:


# diabetes_norm_train_342.txt
# cols [0] to [9] predictors. col [10] target
# norm division constants:
# 100, -1, 100, 1000, 1000, 1000, 100, 10, 10, 1000, 1000
#
0.5900, 1.0000, 0.3210, 0.1010, 0.1570, 0.0932, 0.3800, 0.4000, 0.4860, 0.0870, 0.1510
0.4800, 0.0000, 0.2160, 0.0870, 0.1830, 0.1032, 0.7000, 0.3000, 0.3892, 0.0690, 0.0750
0.7200, 1.0000, 0.3050, 0.0930, 0.1560, 0.0936, 0.4100, 0.4000, 0.4673, 0.0850, 0.1410
0.2400, 0.0000, 0.2530, 0.0840, 0.1980, 0.1314, 0.4000, 0.5000, 0.4890, 0.0890, 0.2060
0.5000, 0.0000, 0.2300, 0.1010, 0.1920, 0.1254, 0.5200, 0.4000, 0.4291, 0.0800, 0.1350
0.2300, 0.0000, 0.2260, 0.0890, 0.1390, 0.0648, 0.6100, 0.2000, 0.4190, 0.0680, 0.0970
0.3600, 1.0000, 0.2200, 0.0900, 0.1600, 0.0996, 0.5000, 0.3000, 0.3951, 0.0820, 0.1380
0.6600, 1.0000, 0.2620, 0.1140, 0.2550, 0.1850, 0.5600, 0.4550, 0.4249, 0.0920, 0.0630
0.6000, 1.0000, 0.3210, 0.0830, 0.1790, 0.1194, 0.4200, 0.4000, 0.4477, 0.0940, 0.1100
0.2900, 0.0000, 0.3000, 0.0850, 0.1800, 0.0934, 0.4300, 0.4000, 0.5385, 0.0880, 0.3100
0.2200, 0.0000, 0.1860, 0.0970, 0.1140, 0.0576, 0.4600, 0.2000, 0.3951, 0.0830, 0.1010
0.5600, 1.0000, 0.2800, 0.0850, 0.1840, 0.1448, 0.3200, 0.6000, 0.3584, 0.0770, 0.0690
0.5300, 0.0000, 0.2370, 0.0920, 0.1860, 0.1092, 0.6200, 0.3000, 0.4304, 0.0810, 0.1790
0.5000, 1.0000, 0.2620, 0.0970, 0.1860, 0.1054, 0.4900, 0.4000, 0.5063, 0.0880, 0.1850
0.6100, 0.0000, 0.2400, 0.0910, 0.2020, 0.1154, 0.7200, 0.3000, 0.4291, 0.0730, 0.1180
0.3400, 1.0000, 0.2470, 0.1180, 0.2540, 0.1842, 0.3900, 0.7000, 0.5037, 0.0810, 0.1710
0.4700, 0.0000, 0.3030, 0.1090, 0.2070, 0.1002, 0.7000, 0.3000, 0.5215, 0.0980, 0.1660
0.6800, 1.0000, 0.2750, 0.1110, 0.2140, 0.1470, 0.3900, 0.5000, 0.4942, 0.0910, 0.1440
0.3800, 0.0000, 0.2540, 0.0840, 0.1620, 0.1030, 0.4200, 0.4000, 0.4443, 0.0870, 0.0970
0.4100, 0.0000, 0.2470, 0.0830, 0.1870, 0.1082, 0.6000, 0.3000, 0.4543, 0.0780, 0.1680
0.3500, 0.0000, 0.2110, 0.0820, 0.1560, 0.0878, 0.5000, 0.3000, 0.4511, 0.0950, 0.0680
0.2500, 1.0000, 0.2430, 0.0950, 0.1620, 0.0986, 0.5400, 0.3000, 0.3850, 0.0870, 0.0490
0.2500, 0.0000, 0.2600, 0.0920, 0.1870, 0.1204, 0.5600, 0.3000, 0.3970, 0.0880, 0.0680
0.6100, 1.0000, 0.3200, 0.1037, 0.2100, 0.0852, 0.3500, 0.6000, 0.6107, 0.1240, 0.2450
0.3100, 0.0000, 0.2970, 0.0880, 0.1670, 0.1034, 0.4800, 0.4000, 0.4357, 0.0780, 0.1840
0.3000, 1.0000, 0.2520, 0.0830, 0.1780, 0.1184, 0.3400, 0.5000, 0.4852, 0.0830, 0.2020
0.1900, 0.0000, 0.1920, 0.0870, 0.1240, 0.0540, 0.5700, 0.2000, 0.4174, 0.0900, 0.1370
0.4200, 0.0000, 0.3190, 0.0830, 0.1580, 0.0876, 0.5300, 0.3000, 0.4466, 0.1010, 0.0850
0.6300, 0.0000, 0.2440, 0.0730, 0.1600, 0.0914, 0.4800, 0.3000, 0.4635, 0.0780, 0.1310
0.6700, 1.0000, 0.2580, 0.1130, 0.1580, 0.0542, 0.6400, 0.2000, 0.5293, 0.1040, 0.2830
0.3200, 0.0000, 0.3050, 0.0890, 0.1820, 0.1106, 0.5600, 0.3000, 0.4344, 0.0890, 0.1290
0.4200, 0.0000, 0.2030, 0.0710, 0.1610, 0.0812, 0.6600, 0.2000, 0.4234, 0.0810, 0.0590
0.5800, 1.0000, 0.3800, 0.1030, 0.1500, 0.1072, 0.2200, 0.7000, 0.4644, 0.0980, 0.3410
0.5700, 0.0000, 0.2170, 0.0940, 0.1570, 0.0580, 0.8200, 0.2000, 0.4443, 0.0920, 0.0870
0.5300, 0.0000, 0.2050, 0.0780, 0.1470, 0.0842, 0.5200, 0.3000, 0.3989, 0.0750, 0.0650
0.6200, 1.0000, 0.2350, 0.0803, 0.2250, 0.1128, 0.8600, 0.2620, 0.4875, 0.0960, 0.1020
0.5200, 0.0000, 0.2850, 0.1100, 0.1950, 0.0972, 0.6000, 0.3000, 0.5242, 0.0850, 0.2650
0.4600, 0.0000, 0.2740, 0.0780, 0.1710, 0.0880, 0.5800, 0.3000, 0.4828, 0.0900, 0.2760
0.4800, 1.0000, 0.3300, 0.1230, 0.2530, 0.1636, 0.4400, 0.6000, 0.5425, 0.0970, 0.2520
0.4800, 1.0000, 0.2770, 0.0730, 0.1910, 0.1194, 0.4600, 0.4000, 0.4852, 0.0920, 0.0900
0.5000, 1.0000, 0.2560, 0.1010, 0.2290, 0.1622, 0.4300, 0.5000, 0.4779, 0.1140, 0.1000
0.2100, 0.0000, 0.2010, 0.0630, 0.1350, 0.0690, 0.5400, 0.3000, 0.4094, 0.0890, 0.0550
0.3200, 1.0000, 0.2540, 0.0903, 0.1530, 0.1004, 0.3400, 0.4500, 0.4533, 0.0830, 0.0610
0.5400, 0.0000, 0.2420, 0.0740, 0.2040, 0.1090, 0.8200, 0.2000, 0.4174, 0.1090, 0.0920
0.6100, 1.0000, 0.3270, 0.0970, 0.1770, 0.1184, 0.2900, 0.6000, 0.4997, 0.0870, 0.2590
0.5600, 1.0000, 0.2310, 0.1040, 0.1810, 0.1164, 0.4700, 0.4000, 0.4477, 0.0790, 0.0530
0.3300, 0.0000, 0.2530, 0.0850, 0.1550, 0.0850, 0.5100, 0.3000, 0.4554, 0.0700, 0.1900
0.2700, 0.0000, 0.1960, 0.0780, 0.1280, 0.0680, 0.4300, 0.3000, 0.4443, 0.0710, 0.1420
0.6700, 1.0000, 0.2250, 0.0980, 0.1910, 0.1192, 0.6100, 0.3000, 0.3989, 0.0860, 0.0750
0.3700, 1.0000, 0.2770, 0.0930, 0.1800, 0.1194, 0.3000, 0.6000, 0.5030, 0.0880, 0.1420
0.5800, 0.0000, 0.2570, 0.0990, 0.1570, 0.0916, 0.4900, 0.3000, 0.4407, 0.0930, 0.1550
0.6500, 1.0000, 0.2790, 0.1030, 0.1590, 0.0968, 0.4200, 0.4000, 0.4615, 0.0860, 0.2250
0.3400, 0.0000, 0.2550, 0.0930, 0.2180, 0.1440, 0.5700, 0.4000, 0.4443, 0.0880, 0.0590
0.4600, 0.0000, 0.2490, 0.1150, 0.1980, 0.1296, 0.5400, 0.4000, 0.4277, 0.1030, 0.1040
0.3500, 0.0000, 0.2870, 0.0970, 0.2040, 0.1268, 0.6400, 0.3000, 0.4190, 0.0930, 0.1820
0.3700, 0.0000, 0.2180, 0.0840, 0.1840, 0.1010, 0.7300, 0.3000, 0.3912, 0.0930, 0.1280
0.3700, 0.0000, 0.3020, 0.0870, 0.1660, 0.0960, 0.4000, 0.4150, 0.5011, 0.0870, 0.0520
0.4100, 0.0000, 0.2050, 0.0800, 0.1240, 0.0488, 0.6400, 0.2000, 0.4025, 0.0750, 0.0370
0.6000, 0.0000, 0.2040, 0.1050, 0.1980, 0.0784, 0.9900, 0.2000, 0.4635, 0.0790, 0.1700
0.6600, 1.0000, 0.2400, 0.0980, 0.2360, 0.1464, 0.5800, 0.4000, 0.5063, 0.0960, 0.1700
0.2900, 0.0000, 0.2600, 0.0830, 0.1410, 0.0652, 0.6400, 0.2000, 0.4078, 0.0830, 0.0610
0.3700, 1.0000, 0.2680, 0.0790, 0.1570, 0.0980, 0.2800, 0.6000, 0.5043, 0.0960, 0.1440
0.4100, 1.0000, 0.2570, 0.0830, 0.1810, 0.1066, 0.6600, 0.3000, 0.3738, 0.0850, 0.0520
0.3900, 0.0000, 0.2290, 0.0770, 0.2040, 0.1432, 0.4600, 0.4000, 0.4304, 0.0740, 0.1280
0.6700, 1.0000, 0.2400, 0.0830, 0.1430, 0.0772, 0.4900, 0.3000, 0.4431, 0.0940, 0.0710
0.3600, 1.0000, 0.2410, 0.1120, 0.1930, 0.1250, 0.3500, 0.6000, 0.5106, 0.0950, 0.1630
0.4600, 1.0000, 0.2470, 0.0850, 0.1740, 0.1232, 0.3000, 0.6000, 0.4644, 0.0960, 0.1500
0.6000, 1.0000, 0.2500, 0.0897, 0.1850, 0.1208, 0.4600, 0.4020, 0.4511, 0.0920, 0.0970
0.5900, 1.0000, 0.2360, 0.0830, 0.1650, 0.1000, 0.4700, 0.4000, 0.4500, 0.0920, 0.1600
0.5300, 0.0000, 0.2210, 0.0930, 0.1340, 0.0762, 0.4600, 0.3000, 0.4078, 0.0960, 0.1780
0.4800, 0.0000, 0.1990, 0.0910, 0.1890, 0.1096, 0.6900, 0.3000, 0.3951, 0.1010, 0.0480
0.4800, 0.0000, 0.2950, 0.1310, 0.2070, 0.1322, 0.4700, 0.4000, 0.4935, 0.1060, 0.2700
0.6600, 1.0000, 0.2600, 0.0910, 0.2640, 0.1466, 0.6500, 0.4000, 0.5568, 0.0870, 0.2020
0.5200, 1.0000, 0.2450, 0.0940, 0.2170, 0.1494, 0.4800, 0.5000, 0.4585, 0.0890, 0.1110
0.5200, 1.0000, 0.2660, 0.1110, 0.2090, 0.1264, 0.6100, 0.3000, 0.4682, 0.1090, 0.0850
0.4600, 1.0000, 0.2350, 0.0870, 0.1810, 0.1148, 0.4400, 0.4000, 0.4710, 0.0980, 0.0420
0.4000, 1.0000, 0.2900, 0.1150, 0.0970, 0.0472, 0.3500, 0.2770, 0.4304, 0.0950, 0.1700
0.2200, 0.0000, 0.2300, 0.0730, 0.1610, 0.0978, 0.5400, 0.3000, 0.3829, 0.0910, 0.2000
0.5000, 0.0000, 0.2100, 0.0880, 0.1400, 0.0718, 0.3500, 0.4000, 0.5112, 0.0710, 0.2520
0.2000, 0.0000, 0.2290, 0.0870, 0.1910, 0.1282, 0.5300, 0.4000, 0.3892, 0.0850, 0.1130
0.6800, 0.0000, 0.2750, 0.1070, 0.2410, 0.1496, 0.6400, 0.4000, 0.4920, 0.0900, 0.1430
0.5200, 1.0000, 0.2430, 0.0860, 0.1970, 0.1336, 0.4400, 0.5000, 0.4575, 0.0910, 0.0510
0.4400, 0.0000, 0.2310, 0.0870, 0.2130, 0.1264, 0.7700, 0.3000, 0.3871, 0.0720, 0.0520
0.3800, 0.0000, 0.2730, 0.0810, 0.1460, 0.0816, 0.4700, 0.3000, 0.4466, 0.0810, 0.2100
0.4900, 0.0000, 0.2270, 0.0653, 0.1680, 0.0962, 0.6200, 0.2710, 0.3892, 0.0600, 0.0650
0.6100, 0.0000, 0.3300, 0.0950, 0.1820, 0.1148, 0.5400, 0.3000, 0.4190, 0.0740, 0.1410
0.2900, 1.0000, 0.1940, 0.0830, 0.1520, 0.1058, 0.3900, 0.4000, 0.3584, 0.0830, 0.0550
0.6100, 0.0000, 0.2580, 0.0980, 0.2350, 0.1258, 0.7600, 0.3000, 0.5112, 0.0820, 0.1340
0.3400, 1.0000, 0.2260, 0.0750, 0.1660, 0.0918, 0.6000, 0.3000, 0.4263, 0.1080, 0.0420
0.3600, 0.0000, 0.2190, 0.0890, 0.1890, 0.1052, 0.6800, 0.3000, 0.4369, 0.0960, 0.1110
0.5200, 0.0000, 0.2400, 0.0830, 0.1670, 0.0866, 0.7100, 0.2000, 0.3850, 0.0940, 0.0980
0.6100, 0.0000, 0.3120, 0.0790, 0.2350, 0.1568, 0.4700, 0.5000, 0.5050, 0.0960, 0.1640
0.4300, 0.0000, 0.2680, 0.1230, 0.1930, 0.1022, 0.6700, 0.3000, 0.4779, 0.0940, 0.0480
0.3500, 0.0000, 0.2040, 0.0650, 0.1870, 0.1056, 0.6700, 0.2790, 0.4277, 0.0780, 0.0960
0.2700, 0.0000, 0.2480, 0.0910, 0.1890, 0.1068, 0.6900, 0.3000, 0.4190, 0.0690, 0.0900
0.2900, 0.0000, 0.2100, 0.0710, 0.1560, 0.0970, 0.3800, 0.4000, 0.4654, 0.0900, 0.1620
0.6400, 1.0000, 0.2730, 0.1090, 0.1860, 0.1076, 0.3800, 0.5000, 0.5308, 0.0990, 0.1500
0.4100, 0.0000, 0.3460, 0.0873, 0.2050, 0.1426, 0.4100, 0.5000, 0.4673, 0.1100, 0.2790
0.4900, 1.0000, 0.2590, 0.0910, 0.1780, 0.1066, 0.5200, 0.3000, 0.4575, 0.0750, 0.0920
0.4800, 0.0000, 0.2040, 0.0980, 0.2090, 0.1394, 0.4600, 0.5000, 0.4771, 0.0780, 0.0830
0.5300, 0.0000, 0.2800, 0.0880, 0.2330, 0.1438, 0.5800, 0.4000, 0.5050, 0.0910, 0.1280
0.5300, 1.0000, 0.2220, 0.1130, 0.1970, 0.1152, 0.6700, 0.3000, 0.4304, 0.1000, 0.1020
0.2300, 0.0000, 0.2900, 0.0900, 0.2160, 0.1314, 0.6500, 0.3000, 0.4585, 0.0910, 0.3020
0.6500, 1.0000, 0.3020, 0.0980, 0.2190, 0.1606, 0.4000, 0.5000, 0.4522, 0.0840, 0.1980
0.4100, 0.0000, 0.3240, 0.0940, 0.1710, 0.1044, 0.5600, 0.3000, 0.3970, 0.0760, 0.0950
0.5500, 1.0000, 0.2340, 0.0830, 0.1660, 0.1016, 0.4600, 0.4000, 0.4522, 0.0960, 0.0530
0.2200, 0.0000, 0.1930, 0.0820, 0.1560, 0.0932, 0.5200, 0.3000, 0.3989, 0.0710, 0.1340
0.5600, 0.0000, 0.3100, 0.0787, 0.1870, 0.1414, 0.3400, 0.5500, 0.4060, 0.0900, 0.1440
0.5400, 1.0000, 0.3060, 0.1033, 0.1440, 0.0798, 0.3000, 0.4800, 0.5142, 0.1010, 0.2320
0.5900, 1.0000, 0.2550, 0.0953, 0.1900, 0.1394, 0.3500, 0.5430, 0.4357, 0.1170, 0.0810
0.6000, 1.0000, 0.2340, 0.0880, 0.1530, 0.0898, 0.5800, 0.3000, 0.3258, 0.0950, 0.1040
0.5400, 0.0000, 0.2680, 0.0870, 0.2060, 0.1220, 0.6800, 0.3000, 0.4382, 0.0800, 0.0590
0.2500, 0.0000, 0.2830, 0.0870, 0.1930, 0.1280, 0.4900, 0.4000, 0.4382, 0.0920, 0.2460
0.5400, 1.0000, 0.2770, 0.1130, 0.2000, 0.1284, 0.3700, 0.5000, 0.5153, 0.1130, 0.2970
0.5500, 0.0000, 0.3660, 0.1130, 0.1990, 0.0944, 0.4300, 0.4630, 0.5730, 0.0970, 0.2580
0.4000, 1.0000, 0.2650, 0.0930, 0.2360, 0.1470, 0.3700, 0.7000, 0.5561, 0.0920, 0.2290
0.6200, 1.0000, 0.3180, 0.1150, 0.1990, 0.1286, 0.4400, 0.5000, 0.4883, 0.0980, 0.2750
0.6500, 0.0000, 0.2440, 0.1200, 0.2220, 0.1356, 0.3700, 0.6000, 0.5509, 0.1240, 0.2810
0.3300, 1.0000, 0.2540, 0.1020, 0.2060, 0.1410, 0.3900, 0.5000, 0.4868, 0.1050, 0.1790
0.5300, 0.0000, 0.2200, 0.0940, 0.1750, 0.0880, 0.5900, 0.3000, 0.4942, 0.0980, 0.2000
0.3500, 0.0000, 0.2680, 0.0980, 0.1620, 0.1036, 0.4500, 0.4000, 0.4205, 0.0860, 0.2000
0.6600, 0.0000, 0.2800, 0.1010, 0.1950, 0.1292, 0.4000, 0.5000, 0.4860, 0.0940, 0.1730
0.6200, 1.0000, 0.3390, 0.1010, 0.2210, 0.1564, 0.3500, 0.6000, 0.4997, 0.1030, 0.1800
0.5000, 1.0000, 0.2960, 0.0943, 0.3000, 0.2424, 0.3300, 0.9090, 0.4812, 0.1090, 0.0840
0.4700, 0.0000, 0.2860, 0.0970, 0.1640, 0.0906, 0.5600, 0.3000, 0.4466, 0.0880, 0.1210
0.4700, 1.0000, 0.2560, 0.0940, 0.1650, 0.0748, 0.4000, 0.4000, 0.5526, 0.0930, 0.1610
0.2400, 0.0000, 0.2070, 0.0870, 0.1490, 0.0806, 0.6100, 0.2000, 0.3611, 0.0780, 0.0990
0.5800, 1.0000, 0.2620, 0.0910, 0.2170, 0.1242, 0.7100, 0.3000, 0.4691, 0.0680, 0.1090
0.3400, 0.0000, 0.2060, 0.0870, 0.1850, 0.1122, 0.5800, 0.3000, 0.4304, 0.0740, 0.1150
0.5100, 0.0000, 0.2790, 0.0960, 0.1960, 0.1222, 0.4200, 0.5000, 0.5069, 0.1200, 0.2680
0.3100, 1.0000, 0.3530, 0.1250, 0.1870, 0.1124, 0.4800, 0.4000, 0.4890, 0.1090, 0.2740
0.2200, 0.0000, 0.1990, 0.0750, 0.1750, 0.1086, 0.5400, 0.3000, 0.4127, 0.0720, 0.1580
0.5300, 1.0000, 0.2440, 0.0920, 0.2140, 0.1460, 0.5000, 0.4000, 0.4500, 0.0970, 0.1070
0.3700, 1.0000, 0.2140, 0.0830, 0.1280, 0.0696, 0.4900, 0.3000, 0.3850, 0.0840, 0.0830
0.2800, 0.0000, 0.3040, 0.0850, 0.1980, 0.1156, 0.6700, 0.3000, 0.4344, 0.0800, 0.1030
0.4700, 0.0000, 0.3160, 0.0840, 0.1540, 0.0880, 0.3000, 0.5100, 0.5199, 0.1050, 0.2720
0.2300, 0.0000, 0.1880, 0.0780, 0.1450, 0.0720, 0.6300, 0.2000, 0.3912, 0.0860, 0.0850
0.5000, 0.0000, 0.3100, 0.1230, 0.1780, 0.1050, 0.4800, 0.4000, 0.4828, 0.0880, 0.2800
0.5800, 1.0000, 0.3670, 0.1170, 0.1660, 0.0938, 0.4400, 0.4000, 0.4949, 0.1090, 0.3360
0.5500, 0.0000, 0.3210, 0.1100, 0.1640, 0.0842, 0.4200, 0.4000, 0.5242, 0.0900, 0.2810
0.6000, 1.0000, 0.2770, 0.1070, 0.1670, 0.1146, 0.3800, 0.4000, 0.4277, 0.0950, 0.1180
0.4100, 0.0000, 0.3080, 0.0810, 0.2140, 0.1520, 0.2800, 0.7600, 0.5136, 0.1230, 0.3170
0.6000, 1.0000, 0.2750, 0.1060, 0.2290, 0.1438, 0.5100, 0.4000, 0.5142, 0.0910, 0.2350
0.4000, 0.0000, 0.2690, 0.0920, 0.2030, 0.1198, 0.7000, 0.3000, 0.4190, 0.0810, 0.0600
0.5700, 1.0000, 0.3070, 0.0900, 0.2040, 0.1478, 0.3400, 0.6000, 0.4710, 0.0930, 0.1740
0.3700, 0.0000, 0.3830, 0.1130, 0.1650, 0.0946, 0.5300, 0.3000, 0.4466, 0.0790, 0.2590
0.4000, 1.0000, 0.3190, 0.0950, 0.1980, 0.1356, 0.3800, 0.5000, 0.4804, 0.0930, 0.1780
0.3300, 0.0000, 0.3500, 0.0890, 0.2000, 0.1304, 0.4200, 0.4760, 0.4927, 0.1010, 0.1280
0.3200, 1.0000, 0.2780, 0.0890, 0.2160, 0.1462, 0.5500, 0.4000, 0.4304, 0.0910, 0.0960
0.3500, 1.0000, 0.2590, 0.0810, 0.1740, 0.1024, 0.3100, 0.6000, 0.5313, 0.0820, 0.1260
0.5500, 0.0000, 0.3290, 0.1020, 0.1640, 0.1062, 0.4100, 0.4000, 0.4431, 0.0890, 0.2880
0.4900, 0.0000, 0.2600, 0.0930, 0.1830, 0.1002, 0.6400, 0.3000, 0.4543, 0.0880, 0.0880
0.3900, 1.0000, 0.2630, 0.1150, 0.2180, 0.1582, 0.3200, 0.7000, 0.4935, 0.1090, 0.2920
0.6000, 1.0000, 0.2230, 0.1130, 0.1860, 0.1258, 0.4600, 0.4000, 0.4263, 0.0940, 0.0710
0.6700, 1.0000, 0.2830, 0.0930, 0.2040, 0.1322, 0.4900, 0.4000, 0.4736, 0.0920, 0.1970
0.4100, 1.0000, 0.3200, 0.1090, 0.2510, 0.1706, 0.4900, 0.5000, 0.5056, 0.1030, 0.1860
0.4400, 0.0000, 0.2540, 0.0950, 0.1620, 0.0926, 0.5300, 0.3000, 0.4407, 0.0830, 0.0250
0.4800, 1.0000, 0.2330, 0.0893, 0.2120, 0.1428, 0.4600, 0.4610, 0.4754, 0.0980, 0.0840
0.4500, 0.0000, 0.2030, 0.0743, 0.1900, 0.1262, 0.4900, 0.3880, 0.4304, 0.0790, 0.0960
0.4700, 0.0000, 0.3040, 0.1200, 0.1990, 0.1200, 0.4600, 0.4000, 0.5106, 0.0870, 0.1950
0.4600, 0.0000, 0.2060, 0.0730, 0.1720, 0.1070, 0.5100, 0.3000, 0.4249, 0.0800, 0.0530
0.3600, 1.0000, 0.3230, 0.1150, 0.2860, 0.1994, 0.3900, 0.7000, 0.5472, 0.1120, 0.2170
0.3400, 0.0000, 0.2920, 0.0730, 0.1720, 0.1082, 0.4900, 0.4000, 0.4304, 0.0910, 0.1720
0.5300, 1.0000, 0.3310, 0.1170, 0.1830, 0.1190, 0.4800, 0.4000, 0.4382, 0.1060, 0.1310
0.6100, 0.0000, 0.2460, 0.1010, 0.2090, 0.1068, 0.7700, 0.3000, 0.4836, 0.0880, 0.2140
0.3700, 0.0000, 0.2020, 0.0810, 0.1620, 0.0878, 0.6300, 0.3000, 0.4025, 0.0880, 0.0590
0.3300, 1.0000, 0.2080, 0.0840, 0.1250, 0.0702, 0.4600, 0.3000, 0.3784, 0.0660, 0.0700
0.6800, 0.0000, 0.3280, 0.1057, 0.2050, 0.1164, 0.4000, 0.5130, 0.5493, 0.1170, 0.2200
0.4900, 1.0000, 0.3190, 0.0940, 0.2340, 0.1558, 0.3400, 0.7000, 0.5398, 0.1220, 0.2680
0.4800, 0.0000, 0.2390, 0.1090, 0.2320, 0.1052, 0.3700, 0.6000, 0.6107, 0.0960, 0.1520
0.5500, 1.0000, 0.2450, 0.0840, 0.1790, 0.1058, 0.6600, 0.3000, 0.3584, 0.0870, 0.0470
0.4300, 0.0000, 0.2210, 0.0660, 0.1340, 0.0772, 0.4500, 0.3000, 0.4078, 0.0800, 0.0740
0.6000, 1.0000, 0.3300, 0.0970, 0.2170, 0.1256, 0.4500, 0.5000, 0.5447, 0.1120, 0.2950
0.3100, 1.0000, 0.1900, 0.0930, 0.1370, 0.0730, 0.4700, 0.3000, 0.4443, 0.0780, 0.1010
0.5300, 1.0000, 0.2730, 0.0820, 0.1190, 0.0550, 0.3900, 0.3000, 0.4828, 0.0930, 0.1510
0.6700, 0.0000, 0.2280, 0.0870, 0.1660, 0.0986, 0.5200, 0.3000, 0.4344, 0.0920, 0.1270
0.6100, 1.0000, 0.2820, 0.1060, 0.2040, 0.1320, 0.5200, 0.4000, 0.4605, 0.0960, 0.2370
0.6200, 0.0000, 0.2890, 0.0873, 0.2060, 0.1272, 0.3300, 0.6240, 0.5434, 0.0990, 0.2250
0.6000, 0.0000, 0.2560, 0.0870, 0.2070, 0.1258, 0.6900, 0.3000, 0.4111, 0.0840, 0.0810
0.4200, 0.0000, 0.2490, 0.0910, 0.2040, 0.1418, 0.3800, 0.5000, 0.4796, 0.0890, 0.1510
0.3800, 1.0000, 0.2680, 0.1050, 0.1810, 0.1192, 0.3700, 0.5000, 0.4820, 0.0910, 0.1070
0.6200, 0.0000, 0.2240, 0.0790, 0.2220, 0.1474, 0.5900, 0.4000, 0.4357, 0.0760, 0.0640
0.6100, 1.0000, 0.2690, 0.1110, 0.2360, 0.1724, 0.3900, 0.6000, 0.4812, 0.0890, 0.1380
0.6100, 1.0000, 0.2310, 0.1130, 0.1860, 0.1144, 0.4700, 0.4000, 0.4812, 0.1050, 0.1850
0.5300, 0.0000, 0.2860, 0.0880, 0.1710, 0.0988, 0.4100, 0.4000, 0.5050, 0.0990, 0.2650
0.2800, 1.0000, 0.2470, 0.0970, 0.1750, 0.0996, 0.3200, 0.5000, 0.5380, 0.0870, 0.1010
0.2600, 1.0000, 0.3030, 0.0890, 0.2180, 0.1522, 0.3100, 0.7000, 0.5159, 0.0820, 0.1370
0.3000, 0.0000, 0.2130, 0.0870, 0.1340, 0.0630, 0.6300, 0.2000, 0.3689, 0.0660, 0.1430
0.5000, 0.0000, 0.2610, 0.1090, 0.2430, 0.1606, 0.6200, 0.4000, 0.4625, 0.0890, 0.1410
0.4800, 0.0000, 0.2020, 0.0950, 0.1870, 0.1174, 0.5300, 0.4000, 0.4419, 0.0850, 0.0790
0.5100, 0.0000, 0.2520, 0.1030, 0.1760, 0.1122, 0.3700, 0.5000, 0.4898, 0.0900, 0.2920
0.4700, 1.0000, 0.2250, 0.0820, 0.1310, 0.0668, 0.4100, 0.3000, 0.4754, 0.0890, 0.1780
0.6400, 1.0000, 0.2350, 0.0970, 0.2030, 0.1290, 0.5900, 0.3000, 0.4318, 0.0770, 0.0910
0.5100, 1.0000, 0.2590, 0.0760, 0.2400, 0.1690, 0.3900, 0.6000, 0.5075, 0.0960, 0.1160
0.3000, 0.0000, 0.2090, 0.1040, 0.1520, 0.0838, 0.4700, 0.3000, 0.4663, 0.0970, 0.0860
0.5600, 1.0000, 0.2870, 0.0990, 0.2080, 0.1464, 0.3900, 0.5000, 0.4727, 0.0970, 0.1220
0.4200, 0.0000, 0.2210, 0.0850, 0.2130, 0.1386, 0.6000, 0.4000, 0.4277, 0.0940, 0.0720
0.6200, 1.0000, 0.2670, 0.1150, 0.1830, 0.1240, 0.3500, 0.5000, 0.4788, 0.1000, 0.1290
0.3400, 0.0000, 0.3140, 0.0870, 0.1490, 0.0938, 0.4600, 0.3000, 0.3829, 0.0770, 0.1420
0.6000, 0.0000, 0.2220, 0.1047, 0.2210, 0.1054, 0.6000, 0.3680, 0.5628, 0.0930, 0.0900
0.6400, 0.0000, 0.2100, 0.0923, 0.2270, 0.1468, 0.6500, 0.3490, 0.4331, 0.1020, 0.1580
0.3900, 1.0000, 0.2120, 0.0900, 0.1820, 0.1104, 0.6000, 0.3000, 0.4060, 0.0980, 0.0390
0.7100, 1.0000, 0.2650, 0.1050, 0.2810, 0.1736, 0.5500, 0.5000, 0.5568, 0.0840, 0.1960
0.4800, 1.0000, 0.2920, 0.1100, 0.2180, 0.1516, 0.3900, 0.6000, 0.4920, 0.0980, 0.2220
0.7900, 1.0000, 0.2700, 0.1030, 0.1690, 0.1108, 0.3700, 0.5000, 0.4663, 0.1100, 0.2770
0.4000, 0.0000, 0.3070, 0.0990, 0.1770, 0.0854, 0.5000, 0.4000, 0.5338, 0.0850, 0.0990
0.4900, 1.0000, 0.2880, 0.0920, 0.2070, 0.1400, 0.4400, 0.5000, 0.4745, 0.0920, 0.1960
0.5100, 0.0000, 0.3060, 0.1030, 0.1980, 0.1066, 0.5700, 0.3000, 0.5148, 0.1000, 0.2020
0.5700, 0.0000, 0.3010, 0.1170, 0.2020, 0.1396, 0.4200, 0.5000, 0.4625, 0.1200, 0.1550
0.5900, 1.0000, 0.2470, 0.1140, 0.1520, 0.1048, 0.2900, 0.5000, 0.4511, 0.0880, 0.0770
0.5100, 0.0000, 0.2770, 0.0990, 0.2290, 0.1456, 0.6900, 0.3000, 0.4277, 0.0770, 0.1910
0.7400, 0.0000, 0.2980, 0.1010, 0.1710, 0.1048, 0.5000, 0.3000, 0.4394, 0.0860, 0.0700
0.6700, 0.0000, 0.2670, 0.1050, 0.2250, 0.1354, 0.6900, 0.3000, 0.4635, 0.0960, 0.0730
0.4900, 0.0000, 0.1980, 0.0880, 0.1880, 0.1148, 0.5700, 0.3000, 0.4394, 0.0930, 0.0490
0.5700, 0.0000, 0.2330, 0.0880, 0.1550, 0.0636, 0.7800, 0.2000, 0.4205, 0.0780, 0.0650
0.5600, 1.0000, 0.3510, 0.1230, 0.1640, 0.0950, 0.3800, 0.4000, 0.5043, 0.1170, 0.2630
0.5200, 1.0000, 0.2970, 0.1090, 0.2280, 0.1628, 0.3100, 0.8000, 0.5142, 0.1030, 0.2480
0.6900, 0.0000, 0.2930, 0.1240, 0.2230, 0.1390, 0.5400, 0.4000, 0.5011, 0.1020, 0.2960
0.3700, 0.0000, 0.2030, 0.0830, 0.1850, 0.1246, 0.3800, 0.5000, 0.4719, 0.0880, 0.2140
0.2400, 0.0000, 0.2250, 0.0890, 0.1410, 0.0680, 0.5200, 0.3000, 0.4654, 0.0840, 0.1850
0.5500, 1.0000, 0.2270, 0.0930, 0.1540, 0.0942, 0.5300, 0.3000, 0.3526, 0.0750, 0.0780
0.3600, 0.0000, 0.2280, 0.0870, 0.1780, 0.1160, 0.4100, 0.4000, 0.4654, 0.0820, 0.0930
0.4200, 1.0000, 0.2400, 0.1070, 0.1500, 0.0850, 0.4400, 0.3000, 0.4654, 0.0960, 0.2520
0.2100, 0.0000, 0.2420, 0.0760, 0.1470, 0.0770, 0.5300, 0.3000, 0.4443, 0.0790, 0.1500
0.4100, 0.0000, 0.2020, 0.0620, 0.1530, 0.0890, 0.5000, 0.3000, 0.4249, 0.0890, 0.0770
0.5700, 1.0000, 0.2940, 0.1090, 0.1600, 0.0876, 0.3100, 0.5000, 0.5333, 0.0920, 0.2080
0.2000, 1.0000, 0.2210, 0.0870, 0.1710, 0.0996, 0.5800, 0.3000, 0.4205, 0.0780, 0.0770
0.6700, 1.0000, 0.2360, 0.1113, 0.1890, 0.1054, 0.7000, 0.2700, 0.4220, 0.0930, 0.1080
0.3400, 0.0000, 0.2520, 0.0770, 0.1890, 0.1206, 0.5300, 0.4000, 0.4344, 0.0790, 0.1600
0.4100, 1.0000, 0.2490, 0.0860, 0.1920, 0.1150, 0.6100, 0.3000, 0.4382, 0.0940, 0.0530
0.3800, 1.0000, 0.3300, 0.0780, 0.3010, 0.2150, 0.5000, 0.6020, 0.5193, 0.1080, 0.2200
0.5100, 0.0000, 0.2350, 0.1010, 0.1950, 0.1210, 0.5100, 0.4000, 0.4745, 0.0940, 0.1540
0.5200, 1.0000, 0.2640, 0.0913, 0.2180, 0.1520, 0.3900, 0.5590, 0.4905, 0.0990, 0.2590
0.6700, 0.0000, 0.2980, 0.0800, 0.1720, 0.0934, 0.6300, 0.3000, 0.4357, 0.0820, 0.0900
0.6100, 0.0000, 0.3000, 0.1080, 0.1940, 0.1000, 0.5200, 0.3730, 0.5347, 0.1050, 0.2460
0.6700, 1.0000, 0.2500, 0.1117, 0.1460, 0.0934, 0.3300, 0.4420, 0.4585, 0.1030, 0.1240
0.5600, 0.0000, 0.2700, 0.1050, 0.2470, 0.1606, 0.5400, 0.5000, 0.5088, 0.0940, 0.0670
0.6400, 0.0000, 0.2000, 0.0747, 0.1890, 0.1148, 0.6200, 0.3050, 0.4111, 0.0910, 0.0720
0.5800, 1.0000, 0.2550, 0.1120, 0.1630, 0.1106, 0.2900, 0.6000, 0.4762, 0.0860, 0.2570
0.5500, 0.0000, 0.2820, 0.0910, 0.2500, 0.1402, 0.6700, 0.4000, 0.5366, 0.1030, 0.2620
0.6200, 1.0000, 0.3330, 0.1140, 0.1820, 0.1140, 0.3800, 0.5000, 0.5011, 0.0960, 0.2750
0.5700, 1.0000, 0.2560, 0.0960, 0.2000, 0.1330, 0.5200, 0.3850, 0.4318, 0.1050, 0.1770
0.2000, 1.0000, 0.2420, 0.0880, 0.1260, 0.0722, 0.4500, 0.3000, 0.3784, 0.0740, 0.0710
0.5300, 1.0000, 0.2210, 0.0980, 0.1650, 0.1052, 0.4700, 0.4000, 0.4159, 0.0810, 0.0470
0.3200, 1.0000, 0.3140, 0.0890, 0.1530, 0.0842, 0.5600, 0.3000, 0.4159, 0.0900, 0.1870
0.4100, 0.0000, 0.2310, 0.0860, 0.1480, 0.0780, 0.5800, 0.3000, 0.4094, 0.0600, 0.1250
0.6000, 0.0000, 0.2340, 0.0767, 0.2470, 0.1480, 0.6500, 0.3800, 0.5136, 0.0770, 0.0780
0.2600, 0.0000, 0.1880, 0.0830, 0.1910, 0.1036, 0.6900, 0.3000, 0.4522, 0.0690, 0.0510
0.3700, 0.0000, 0.3080, 0.1120, 0.2820, 0.1972, 0.4300, 0.7000, 0.5342, 0.1010, 0.2580
0.4500, 0.0000, 0.3200, 0.1100, 0.2240, 0.1342, 0.4500, 0.5000, 0.5412, 0.0930, 0.2150
0.6700, 0.0000, 0.3160, 0.1160, 0.1790, 0.0904, 0.4100, 0.4000, 0.5472, 0.1000, 0.3030
0.3400, 1.0000, 0.3550, 0.1200, 0.2330, 0.1466, 0.3400, 0.7000, 0.5568, 0.1010, 0.2430
0.5000, 0.0000, 0.3190, 0.0783, 0.2070, 0.1492, 0.3800, 0.5450, 0.4595, 0.0840, 0.0910
0.7100, 0.0000, 0.2950, 0.0970, 0.2270, 0.1516, 0.4500, 0.5000, 0.5024, 0.1080, 0.1500
0.5700, 1.0000, 0.3160, 0.1170, 0.2250, 0.1076, 0.4000, 0.6000, 0.5958, 0.1130, 0.3100
0.4900, 0.0000, 0.2030, 0.0930, 0.1840, 0.1030, 0.6100, 0.3000, 0.4605, 0.0930, 0.1530
0.3500, 0.0000, 0.4130, 0.0810, 0.1680, 0.1028, 0.3700, 0.5000, 0.4949, 0.0940, 0.3460
0.4100, 1.0000, 0.2120, 0.1020, 0.1840, 0.1004, 0.6400, 0.3000, 0.4585, 0.0790, 0.0630
0.7000, 1.0000, 0.2410, 0.0823, 0.1940, 0.1492, 0.3100, 0.6260, 0.4234, 0.1050, 0.0890
0.5200, 0.0000, 0.2300, 0.1070, 0.1790, 0.1237, 0.4250, 0.4210, 0.4159, 0.0930, 0.0500
0.6000, 0.0000, 0.2560, 0.0780, 0.1950, 0.0954, 0.9100, 0.2000, 0.3761, 0.0870, 0.0390
0.6200, 0.0000, 0.2250, 0.1250, 0.2150, 0.0990, 0.9800, 0.2000, 0.4500, 0.0950, 0.1030
0.4400, 1.0000, 0.3820, 0.1230, 0.2010, 0.1266, 0.4400, 0.5000, 0.5024, 0.0920, 0.3080
0.2800, 1.0000, 0.1920, 0.0810, 0.1550, 0.0946, 0.5100, 0.3000, 0.3850, 0.0870, 0.1160
0.5800, 1.0000, 0.2900, 0.0850, 0.1560, 0.1092, 0.3600, 0.4000, 0.3989, 0.0860, 0.1450
0.3900, 1.0000, 0.2400, 0.0897, 0.1900, 0.1136, 0.5200, 0.3650, 0.4804, 0.1010, 0.0740
0.3400, 1.0000, 0.2060, 0.0980, 0.1830, 0.0920, 0.8300, 0.2000, 0.3689, 0.0920, 0.0450
0.6500, 0.0000, 0.2630, 0.0700, 0.2440, 0.1662, 0.5100, 0.5000, 0.4898, 0.0980, 0.1150
0.6600, 1.0000, 0.3460, 0.1150, 0.2040, 0.1394, 0.3600, 0.6000, 0.4963, 0.1090, 0.2640
0.5100, 0.0000, 0.2340, 0.0870, 0.2200, 0.1088, 0.9300, 0.2000, 0.4511, 0.0820, 0.0870
0.5000, 1.0000, 0.2920, 0.1190, 0.1620, 0.0852, 0.5400, 0.3000, 0.4736, 0.0950, 0.2020
0.5900, 1.0000, 0.2720, 0.1070, 0.1580, 0.1020, 0.3900, 0.4000, 0.4443, 0.0930, 0.1270
0.5200, 0.0000, 0.2700, 0.0783, 0.1340, 0.0730, 0.4400, 0.3050, 0.4443, 0.0690, 0.1820
0.6900, 1.0000, 0.2450, 0.1080, 0.2430, 0.1364, 0.4000, 0.6000, 0.5808, 0.1000, 0.2410
0.5300, 0.0000, 0.2410, 0.1050, 0.1840, 0.1134, 0.4600, 0.4000, 0.4812, 0.0950, 0.0660
0.4700, 1.0000, 0.2530, 0.0980, 0.1730, 0.1056, 0.4400, 0.4000, 0.4762, 0.1080, 0.0940
0.5200, 0.0000, 0.2880, 0.1130, 0.2800, 0.1740, 0.6700, 0.4000, 0.5273, 0.0860, 0.2830
0.3900, 0.0000, 0.2090, 0.0950, 0.1500, 0.0656, 0.6800, 0.2000, 0.4407, 0.0950, 0.0640
0.6700, 1.0000, 0.2300, 0.0700, 0.1840, 0.1280, 0.3500, 0.5000, 0.4654, 0.0990, 0.1020
0.5900, 1.0000, 0.2410, 0.0960, 0.1700, 0.0986, 0.5400, 0.3000, 0.4466, 0.0850, 0.2000
0.5100, 1.0000, 0.2810, 0.1060, 0.2020, 0.1222, 0.5500, 0.4000, 0.4820, 0.0870, 0.2650
0.2300, 1.0000, 0.1800, 0.0780, 0.1710, 0.0960, 0.4800, 0.4000, 0.4905, 0.0920, 0.0940
0.6800, 0.0000, 0.2590, 0.0930, 0.2530, 0.1812, 0.5300, 0.5000, 0.4543, 0.0980, 0.2300
0.4400, 0.0000, 0.2150, 0.0850, 0.1570, 0.0922, 0.5500, 0.3000, 0.3892, 0.0840, 0.1810
0.6000, 1.0000, 0.2430, 0.1030, 0.1410, 0.0866, 0.3300, 0.4000, 0.4673, 0.0780, 0.1560
0.5200, 0.0000, 0.2450, 0.0900, 0.1980, 0.1290, 0.2900, 0.7000, 0.5298, 0.0860, 0.2330
0.3800, 0.0000, 0.2130, 0.0720, 0.1650, 0.0602, 0.8800, 0.2000, 0.4431, 0.0900, 0.0600
0.6100, 0.0000, 0.2580, 0.0900, 0.2800, 0.1954, 0.5500, 0.5000, 0.4997, 0.0900, 0.2190
0.6800, 1.0000, 0.2480, 0.1010, 0.2210, 0.1514, 0.6000, 0.4000, 0.3871, 0.0870, 0.0800
0.2800, 1.0000, 0.3150, 0.0830, 0.2280, 0.1494, 0.3800, 0.6000, 0.5313, 0.0830, 0.0680
0.6500, 1.0000, 0.3350, 0.1020, 0.1900, 0.1262, 0.3500, 0.5000, 0.4970, 0.1020, 0.3320
0.6900, 0.0000, 0.2810, 0.1130, 0.2340, 0.1428, 0.5200, 0.4000, 0.5278, 0.0770, 0.2480
0.5100, 0.0000, 0.2430, 0.0853, 0.1530, 0.0716, 0.7100, 0.2150, 0.3951, 0.0820, 0.0840
0.2900, 0.0000, 0.3500, 0.0983, 0.2040, 0.1426, 0.5000, 0.4080, 0.4043, 0.0910, 0.2000
0.5500, 1.0000, 0.2350, 0.0930, 0.1770, 0.1268, 0.4100, 0.4000, 0.3829, 0.0830, 0.0550
0.3400, 1.0000, 0.3000, 0.0830, 0.1850, 0.1072, 0.5300, 0.3000, 0.4820, 0.0920, 0.0850
0.6700, 0.0000, 0.2070, 0.0830, 0.1700, 0.0998, 0.5900, 0.3000, 0.4025, 0.0770, 0.0890
0.4900, 0.0000, 0.2560, 0.0760, 0.1610, 0.0998, 0.5100, 0.3000, 0.3932, 0.0780, 0.0310
0.5500, 1.0000, 0.2290, 0.0810, 0.1230, 0.0672, 0.4100, 0.3000, 0.4304, 0.0880, 0.1290
0.5900, 1.0000, 0.2510, 0.0900, 0.1630, 0.1014, 0.4600, 0.4000, 0.4357, 0.0910, 0.0830
0.5300, 0.0000, 0.3320, 0.0827, 0.1860, 0.1068, 0.4600, 0.4040, 0.5112, 0.1020, 0.2750
0.4800, 1.0000, 0.2410, 0.1100, 0.2090, 0.1346, 0.5800, 0.4000, 0.4407, 0.1000, 0.0650
0.5200, 0.0000, 0.2950, 0.1043, 0.2110, 0.1328, 0.4900, 0.4310, 0.4984, 0.0980, 0.1980
0.6900, 0.0000, 0.2960, 0.1220, 0.2310, 0.1284, 0.5600, 0.4000, 0.5451, 0.0860, 0.2360
0.6000, 1.0000, 0.2280, 0.1100, 0.2450, 0.1898, 0.3900, 0.6000, 0.4394, 0.0880, 0.2530
0.4600, 1.0000, 0.2270, 0.0830, 0.1830, 0.1258, 0.3200, 0.6000, 0.4836, 0.0750, 0.1240
0.5100, 1.0000, 0.2620, 0.1010, 0.1610, 0.0996, 0.4800, 0.3000, 0.4205, 0.0880, 0.0440
0.6700, 1.0000, 0.2350, 0.0960, 0.2070, 0.1382, 0.4200, 0.5000, 0.4898, 0.1110, 0.1720
0.4900, 0.0000, 0.2210, 0.0850, 0.1360, 0.0634, 0.6200, 0.2190, 0.3970, 0.0720, 0.1140
0.4600, 1.0000, 0.2650, 0.0940, 0.2470, 0.1602, 0.5900, 0.4000, 0.4935, 0.1110, 0.1420
0.4700, 0.0000, 0.3240, 0.1050, 0.1880, 0.1250, 0.4600, 0.4090, 0.4443, 0.0990, 0.1090
0.7500, 0.0000, 0.3010, 0.0780, 0.2220, 0.1542, 0.4400, 0.5050, 0.4779, 0.0970, 0.1800
0.2800, 0.0000, 0.2420, 0.0930, 0.1740, 0.1064, 0.5400, 0.3000, 0.4220, 0.0840, 0.1440
0.6500, 1.0000, 0.3130, 0.1100, 0.2130, 0.1280, 0.4700, 0.5000, 0.5247, 0.0910, 0.1630
0.4200, 0.0000, 0.3010, 0.0910, 0.1820, 0.1148, 0.4900, 0.4000, 0.4511, 0.0820, 0.1470
0.5100, 0.0000, 0.2450, 0.0790, 0.2120, 0.1286, 0.6500, 0.3000, 0.4522, 0.0910, 0.0970
0.5300, 1.0000, 0.2770, 0.0950, 0.1900, 0.1018, 0.4100, 0.5000, 0.5464, 0.1010, 0.2200
0.5400, 0.0000, 0.2320, 0.1107, 0.2380, 0.1628, 0.4800, 0.4960, 0.4913, 0.1080, 0.1900
0.7300, 0.0000, 0.2700, 0.1020, 0.2110, 0.1210, 0.6700, 0.3000, 0.4745, 0.0990, 0.1090
0.5400, 0.0000, 0.2680, 0.1080, 0.1760, 0.0806, 0.6700, 0.3000, 0.4956, 0.1060, 0.1910
0.4200, 0.0000, 0.2920, 0.0930, 0.2490, 0.1742, 0.4500, 0.6000, 0.5004, 0.0920, 0.1220
0.7500, 0.0000, 0.3120, 0.1177, 0.2290, 0.1388, 0.2900, 0.7900, 0.5724, 0.1060, 0.2300
0.5500, 1.0000, 0.3210, 0.1127, 0.2070, 0.0924, 0.2500, 0.8280, 0.6105, 0.1110, 0.2420
0.6800, 1.0000, 0.2570, 0.1090, 0.2330, 0.1126, 0.3500, 0.7000, 0.6057, 0.1050, 0.2480
0.5700, 0.0000, 0.2690, 0.0980, 0.2460, 0.1652, 0.3800, 0.7000, 0.5366, 0.0960, 0.2490
0.4800, 0.0000, 0.3140, 0.0753, 0.2420, 0.1516, 0.3800, 0.6370, 0.5568, 0.1030, 0.1920
0.6100, 1.0000, 0.2560, 0.0850, 0.1840, 0.1162, 0.3900, 0.5000, 0.4970, 0.0980, 0.1310
0.6900, 0.0000, 0.3700, 0.1030, 0.2070, 0.1314, 0.5500, 0.4000, 0.4635, 0.0900, 0.2370
0.3800, 0.0000, 0.3260, 0.0770, 0.1680, 0.1006, 0.4700, 0.4000, 0.4625, 0.0960, 0.0780
0.4500, 1.0000, 0.2120, 0.0940, 0.1690, 0.0968, 0.5500, 0.3000, 0.4454, 0.1020, 0.1350
0.5100, 1.0000, 0.2920, 0.1070, 0.1870, 0.1390, 0.3200, 0.6000, 0.4382, 0.0950, 0.2440
0.7100, 1.0000, 0.2400, 0.0840, 0.1380, 0.0858, 0.3900, 0.4000, 0.4190, 0.0900, 0.1990
0.5700, 0.0000, 0.3610, 0.1170, 0.1810, 0.1082, 0.3400, 0.5000, 0.5268, 0.1000, 0.2700
0.5600, 1.0000, 0.2580, 0.1030, 0.1770, 0.1144, 0.3400, 0.5000, 0.4963, 0.0990, 0.1640
0.3200, 1.0000, 0.2200, 0.0880, 0.1370, 0.0786, 0.4800, 0.3000, 0.3951, 0.0780, 0.0720
0.5000, 0.0000, 0.2190, 0.0910, 0.1900, 0.1112, 0.6700, 0.3000, 0.4078, 0.0770, 0.0960
0.4300, 0.0000, 0.3430, 0.0840, 0.2560, 0.1726, 0.3300, 0.8000, 0.5529, 0.1040, 0.3060
0.5400, 1.0000, 0.2520, 0.1150, 0.1810, 0.1200, 0.3900, 0.5000, 0.4701, 0.0920, 0.0910
0.3100, 0.0000, 0.2330, 0.0850, 0.1900, 0.1308, 0.4300, 0.4000, 0.4394, 0.0770, 0.2140
0.5600, 0.0000, 0.2570, 0.0800, 0.2440, 0.1516, 0.5900, 0.4000, 0.5118, 0.0950, 0.0950
0.4400, 0.0000, 0.2510, 0.1330, 0.1820, 0.1130, 0.5500, 0.3000, 0.4249, 0.0840, 0.2160
0.5700, 1.0000, 0.3190, 0.1110, 0.1730, 0.1162, 0.4100, 0.4000, 0.4369, 0.0870, 0.2630

Test data:


# diabetes_norm_test_100.txt
#
0.6400, 1.0000, 0.2840, 0.1110, 0.1840, 0.1270, 0.4100, 0.4000, 0.4382, 0.0970, 0.1780
0.4300, 0.0000, 0.2810, 0.1210, 0.1920, 0.1210, 0.6000, 0.3000, 0.4007, 0.0930, 0.1130
0.1900, 0.0000, 0.2530, 0.0830, 0.2250, 0.1566, 0.4600, 0.5000, 0.4719, 0.0840, 0.2000
0.7100, 1.0000, 0.2610, 0.0850, 0.2200, 0.1524, 0.4700, 0.5000, 0.4635, 0.0910, 0.1390
0.5000, 1.0000, 0.2800, 0.1040, 0.2820, 0.1968, 0.4400, 0.6000, 0.5328, 0.0950, 0.1390
0.5900, 1.0000, 0.2360, 0.0730, 0.1800, 0.1074, 0.5100, 0.4000, 0.4682, 0.0840, 0.0880
0.5700, 0.0000, 0.2450, 0.0930, 0.1860, 0.0966, 0.7100, 0.3000, 0.4522, 0.0910, 0.1480
0.4900, 1.0000, 0.2100, 0.0820, 0.1190, 0.0854, 0.2300, 0.5000, 0.3970, 0.0740, 0.0880
0.4100, 1.0000, 0.3200, 0.1260, 0.1980, 0.1042, 0.4900, 0.4000, 0.5412, 0.1240, 0.2430
0.2500, 1.0000, 0.2260, 0.0850, 0.1300, 0.0710, 0.4800, 0.3000, 0.4007, 0.0810, 0.0710
0.5200, 1.0000, 0.1970, 0.0810, 0.1520, 0.0534, 0.8200, 0.2000, 0.4419, 0.0820, 0.0770
0.3400, 0.0000, 0.2120, 0.0840, 0.2540, 0.1134, 0.5200, 0.5000, 0.6094, 0.0920, 0.1090
0.4200, 1.0000, 0.3060, 0.1010, 0.2690, 0.1722, 0.5000, 0.5000, 0.5455, 0.1060, 0.2720
0.2800, 1.0000, 0.2550, 0.0990, 0.1620, 0.1016, 0.4600, 0.4000, 0.4277, 0.0940, 0.0600
0.4700, 1.0000, 0.2330, 0.0900, 0.1950, 0.1258, 0.5400, 0.4000, 0.4331, 0.0730, 0.0540
0.3200, 1.0000, 0.3100, 0.1000, 0.1770, 0.0962, 0.4500, 0.4000, 0.5187, 0.0770, 0.2210
0.4300, 0.0000, 0.1850, 0.0870, 0.1630, 0.0936, 0.6100, 0.2670, 0.3738, 0.0800, 0.0900
0.5900, 1.0000, 0.2690, 0.1040, 0.1940, 0.1266, 0.4300, 0.5000, 0.4804, 0.1060, 0.3110
0.5300, 0.0000, 0.2830, 0.1010, 0.1790, 0.1070, 0.4800, 0.4000, 0.4788, 0.1010, 0.2810
0.6000, 0.0000, 0.2570, 0.1030, 0.1580, 0.0846, 0.6400, 0.2000, 0.3850, 0.0970, 0.1820
0.5400, 1.0000, 0.3610, 0.1150, 0.1630, 0.0984, 0.4300, 0.4000, 0.4682, 0.1010, 0.3210
0.3500, 1.0000, 0.2410, 0.0947, 0.1550, 0.0974, 0.3200, 0.4840, 0.4852, 0.0940, 0.0580
0.4900, 1.0000, 0.2580, 0.0890, 0.1820, 0.1186, 0.3900, 0.5000, 0.4804, 0.1150, 0.2620
0.5800, 0.0000, 0.2280, 0.0910, 0.1960, 0.1188, 0.4800, 0.4000, 0.4984, 0.1150, 0.2060
0.3600, 1.0000, 0.3910, 0.0900, 0.2190, 0.1358, 0.3800, 0.6000, 0.5421, 0.1030, 0.2330
0.4600, 1.0000, 0.4220, 0.0990, 0.2110, 0.1370, 0.4400, 0.5000, 0.5011, 0.0990, 0.2420
0.4400, 1.0000, 0.2660, 0.0990, 0.2050, 0.1090, 0.4300, 0.5000, 0.5580, 0.1110, 0.1230
0.4600, 0.0000, 0.2990, 0.0830, 0.1710, 0.1130, 0.3800, 0.4500, 0.4585, 0.0980, 0.1670
0.5400, 0.0000, 0.2100, 0.0780, 0.1880, 0.1074, 0.7000, 0.3000, 0.3970, 0.0730, 0.0630
0.6300, 1.0000, 0.2550, 0.1090, 0.2260, 0.1032, 0.4600, 0.5000, 0.5951, 0.0870, 0.1970
0.4100, 1.0000, 0.2420, 0.0900, 0.1990, 0.1236, 0.5700, 0.4000, 0.4522, 0.0860, 0.0710
0.2800, 0.0000, 0.2540, 0.0930, 0.1410, 0.0790, 0.4900, 0.3000, 0.4174, 0.0910, 0.1680
0.1900, 0.0000, 0.2320, 0.0750, 0.1430, 0.0704, 0.5200, 0.3000, 0.4635, 0.0720, 0.1400
0.6100, 1.0000, 0.2610, 0.1260, 0.2150, 0.1298, 0.5700, 0.4000, 0.4949, 0.0960, 0.2170
0.4800, 0.0000, 0.3270, 0.0930, 0.2760, 0.1986, 0.4300, 0.6420, 0.5148, 0.0910, 0.1210
0.5400, 1.0000, 0.2730, 0.1000, 0.2000, 0.1440, 0.3300, 0.6000, 0.4745, 0.0760, 0.2350
0.5300, 1.0000, 0.2660, 0.0930, 0.1850, 0.1224, 0.3600, 0.5000, 0.4890, 0.0820, 0.2450
0.4800, 0.0000, 0.2280, 0.1010, 0.1100, 0.0416, 0.5600, 0.2000, 0.4127, 0.0970, 0.0400
0.5300, 0.0000, 0.2880, 0.1117, 0.1450, 0.0872, 0.4600, 0.3150, 0.4078, 0.0850, 0.0520
0.2900, 1.0000, 0.1810, 0.0730, 0.1580, 0.0990, 0.4100, 0.4000, 0.4500, 0.0780, 0.1040
0.6200, 0.0000, 0.3200, 0.0880, 0.1720, 0.0690, 0.3800, 0.4000, 0.5784, 0.1000, 0.1320
0.5000, 1.0000, 0.2370, 0.0920, 0.1660, 0.0970, 0.5200, 0.3000, 0.4443, 0.0930, 0.0880
0.5800, 1.0000, 0.2360, 0.0960, 0.2570, 0.1710, 0.5900, 0.4000, 0.4905, 0.0820, 0.0690
0.5500, 1.0000, 0.2460, 0.1090, 0.1430, 0.0764, 0.5100, 0.3000, 0.4357, 0.0880, 0.2190
0.5400, 0.0000, 0.2260, 0.0900, 0.1830, 0.1042, 0.6400, 0.3000, 0.4304, 0.0920, 0.0720
0.3600, 0.0000, 0.2780, 0.0730, 0.1530, 0.1044, 0.4200, 0.4000, 0.3497, 0.0730, 0.2010
0.6300, 1.0000, 0.2410, 0.1110, 0.1840, 0.1122, 0.4400, 0.4000, 0.4935, 0.0820, 0.1100
0.4700, 1.0000, 0.2650, 0.0700, 0.1810, 0.1048, 0.6300, 0.3000, 0.4190, 0.0700, 0.0510
0.5100, 1.0000, 0.3280, 0.1120, 0.2020, 0.1006, 0.3700, 0.5000, 0.5775, 0.1090, 0.2770
0.4200, 0.0000, 0.1990, 0.0760, 0.1460, 0.0832, 0.5500, 0.3000, 0.3664, 0.0790, 0.0630
0.3700, 1.0000, 0.2360, 0.0940, 0.2050, 0.1388, 0.5300, 0.4000, 0.4190, 0.1070, 0.1180
0.2800, 0.0000, 0.2210, 0.0820, 0.1680, 0.1006, 0.5400, 0.3000, 0.4205, 0.0860, 0.0690
0.5800, 0.0000, 0.2810, 0.1110, 0.1980, 0.0806, 0.3100, 0.6000, 0.6068, 0.0930, 0.2730
0.3200, 0.0000, 0.2650, 0.0860, 0.1840, 0.1016, 0.5300, 0.4000, 0.4990, 0.0780, 0.2580
0.2500, 1.0000, 0.2350, 0.0880, 0.1430, 0.0808, 0.5500, 0.3000, 0.3584, 0.0830, 0.0430
0.6300, 0.0000, 0.2600, 0.0857, 0.1550, 0.0782, 0.4600, 0.3370, 0.5037, 0.0970, 0.1980
0.5200, 0.0000, 0.2780, 0.0850, 0.2190, 0.1360, 0.4900, 0.4000, 0.5136, 0.0750, 0.2420
0.6500, 1.0000, 0.2850, 0.1090, 0.2010, 0.1230, 0.4600, 0.4000, 0.5075, 0.0960, 0.2320
0.4200, 0.0000, 0.3060, 0.1210, 0.1760, 0.0928, 0.6900, 0.3000, 0.4263, 0.0890, 0.1750
0.5300, 0.0000, 0.2220, 0.0780, 0.1640, 0.0810, 0.7000, 0.2000, 0.4174, 0.1010, 0.0930
0.7900, 1.0000, 0.2330, 0.0880, 0.1860, 0.1284, 0.3300, 0.6000, 0.4812, 0.1020, 0.1680
0.4300, 0.0000, 0.3540, 0.0930, 0.1850, 0.1002, 0.4400, 0.4000, 0.5318, 0.1010, 0.2750
0.4400, 0.0000, 0.3140, 0.1150, 0.1650, 0.0976, 0.5200, 0.3000, 0.4344, 0.0890, 0.2930
0.6200, 1.0000, 0.3780, 0.1190, 0.1130, 0.0510, 0.3100, 0.4000, 0.5043, 0.0840, 0.2810
0.3300, 0.0000, 0.1890, 0.0700, 0.1620, 0.0918, 0.5900, 0.3000, 0.4025, 0.0580, 0.0720
0.5600, 0.0000, 0.3500, 0.0793, 0.1950, 0.1408, 0.4200, 0.4640, 0.4111, 0.0960, 0.1400
0.6600, 0.0000, 0.2170, 0.1260, 0.2120, 0.1278, 0.4500, 0.4710, 0.5278, 0.1010, 0.1890
0.3400, 1.0000, 0.2530, 0.1110, 0.2300, 0.1620, 0.3900, 0.6000, 0.4977, 0.0900, 0.1810
0.4600, 1.0000, 0.2380, 0.0970, 0.2240, 0.1392, 0.4200, 0.5000, 0.5366, 0.0810, 0.2090
0.5000, 0.0000, 0.3180, 0.0820, 0.1360, 0.0692, 0.5500, 0.2000, 0.4078, 0.0850, 0.1360
0.6900, 0.0000, 0.3430, 0.1130, 0.2000, 0.1238, 0.5400, 0.4000, 0.4710, 0.1120, 0.2610
0.3400, 0.0000, 0.2630, 0.0870, 0.1970, 0.1200, 0.6300, 0.3000, 0.4249, 0.0960, 0.1130
0.7100, 1.0000, 0.2700, 0.0933, 0.2690, 0.1902, 0.4100, 0.6560, 0.5242, 0.0930, 0.1310
0.4700, 0.0000, 0.2720, 0.0800, 0.2080, 0.1456, 0.3800, 0.6000, 0.4804, 0.0920, 0.1740
0.4100, 0.0000, 0.3380, 0.1233, 0.1870, 0.1270, 0.4500, 0.4160, 0.4318, 0.1000, 0.2570
0.3400, 0.0000, 0.3300, 0.0730, 0.1780, 0.1146, 0.5100, 0.3490, 0.4127, 0.0920, 0.0550
0.5100, 0.0000, 0.2410, 0.0870, 0.2610, 0.1756, 0.6900, 0.4000, 0.4407, 0.0930, 0.0840
0.4300, 0.0000, 0.2130, 0.0790, 0.1410, 0.0788, 0.5300, 0.3000, 0.3829, 0.0900, 0.0420
0.5500, 0.0000, 0.2300, 0.0947, 0.1900, 0.1376, 0.3800, 0.5000, 0.4277, 0.1060, 0.1460
0.5900, 1.0000, 0.2790, 0.1010, 0.2180, 0.1442, 0.3800, 0.6000, 0.5187, 0.0950, 0.2120
0.2700, 1.0000, 0.3360, 0.1100, 0.2460, 0.1566, 0.5700, 0.4000, 0.5088, 0.0890, 0.2330
0.5100, 1.0000, 0.2270, 0.1030, 0.2170, 0.1624, 0.3000, 0.7000, 0.4812, 0.0800, 0.0910
0.4900, 1.0000, 0.2740, 0.0890, 0.1770, 0.1130, 0.3700, 0.5000, 0.4905, 0.0970, 0.1110
0.2700, 0.0000, 0.2260, 0.0710, 0.1160, 0.0434, 0.5600, 0.2000, 0.4419, 0.0790, 0.1520
0.5700, 1.0000, 0.2320, 0.1073, 0.2310, 0.1594, 0.4100, 0.5630, 0.5030, 0.1120, 0.1200
0.3900, 1.0000, 0.2690, 0.0930, 0.1360, 0.0754, 0.4800, 0.3000, 0.4143, 0.0990, 0.0670
0.6200, 1.0000, 0.3460, 0.1200, 0.2150, 0.1292, 0.4300, 0.5000, 0.5366, 0.1230, 0.3100
0.3700, 0.0000, 0.2330, 0.0880, 0.2230, 0.1420, 0.6500, 0.3400, 0.4357, 0.0820, 0.0940
0.4600, 0.0000, 0.2110, 0.0800, 0.2050, 0.1444, 0.4200, 0.5000, 0.4533, 0.0870, 0.1830
0.6800, 1.0000, 0.2350, 0.1010, 0.1620, 0.0854, 0.5900, 0.3000, 0.4477, 0.0910, 0.0660
0.5100, 0.0000, 0.3150, 0.0930, 0.2310, 0.1440, 0.4900, 0.4700, 0.5252, 0.1170, 0.1730
0.4100, 0.0000, 0.2080, 0.0860, 0.2230, 0.1282, 0.8300, 0.3000, 0.4078, 0.0890, 0.0720
0.5300, 0.0000, 0.2650, 0.0970, 0.1930, 0.1224, 0.5800, 0.3000, 0.4143, 0.0990, 0.0490
0.4500, 0.0000, 0.2420, 0.0830, 0.1770, 0.1184, 0.4500, 0.4000, 0.4220, 0.0820, 0.0640
0.3300, 0.0000, 0.1950, 0.0800, 0.1710, 0.0854, 0.7500, 0.2000, 0.3970, 0.0800, 0.0480
0.6000, 1.0000, 0.2820, 0.1120, 0.1850, 0.1138, 0.4200, 0.4000, 0.4984, 0.0930, 0.1780
0.4700, 1.0000, 0.2490, 0.0750, 0.2250, 0.1660, 0.4200, 0.5000, 0.4443, 0.1020, 0.1040
0.6000, 1.0000, 0.2490, 0.0997, 0.1620, 0.1066, 0.4300, 0.3770, 0.4127, 0.0950, 0.1320
0.3600, 0.0000, 0.3000, 0.0950, 0.2010, 0.1252, 0.4200, 0.4790, 0.5130, 0.0850, 0.2200
0.3600, 0.0000, 0.1960, 0.0710, 0.2500, 0.1332, 0.9700, 0.3000, 0.4595, 0.0920, 0.0570
Posted in Machine Learning, Scikit | Leave a comment

Implementing Ridge Regression from Scratch Python in the Scikit Style Using SAG Training

Posted on June 16, 2026 by jamesdmccaffrey

Ridge regression, aka linear ridge regression, is just linear regression with L2 regularization integrated. The idea is to discourage large model weights, which is a cause of model overfitting. Note: linear ridge regression is sometimes confused with kernel ridge regression, but the two techniques are very, very different.

Basic ridge regression is quite simple. It can be implemented from scratch, with SGD training, in only a few dozen lines of code.

I took a look at the Python scikit library Ridge module. I was extremely surprised at how complicated the module is. First, it uses a very complex variation of SGD training called SAG (stochastic average gradient). And it uses a very tricky algorithm to automatically determine a learning rate. And it uses a non-obvious technique when implementing the L2 alpha constant. And it uses a non-trivial algorithm to auto-stop the training iteration. When I looked at the scikit source code, it was spread over many files and hundreds (more likely thousands — I didn’t try to count) of lines of code.

For reasons that aren’t clear to me, I decided to implement an approximation to the scikit Ridge module, from scratch using Python. I was mostly interested in gaining an understanding of the SAG algorithm. Bottom line: After far too many hours spent on the investigation, I didn’t see any significant advantage of SAG over the much simpler SGD.

SAG is kind of a cross between online (one line of training data at a time for a single gradient) SGD and batch (process all lines to compute an overall gradient) SGD. The SAG algorithm processes one line of data at a time, but stores computed gradients in memory, and updates model weights using a running average gradient.

I used the scikit approach to use the L2 alpha constant. Briefly, the key trick is to scale alpha by dividing by the number of training items.

I used the scikit approach for early stopping for training. Briefly, the training exits if max_change_in_wts / max_weights less-than tolerance (typically 0.0001 or 0.00001). Simple idea, but it was tricky to implement.

I explored the automatic setting of the learning rate, and after a lot of digging through the scikit source code, I figured it out. But I am moderately skeptical that the technique is useful — I think parameterizing the learning rate is a better approach. However, that is a very long and complicated topic which is beyond this blog post.

The output of a demo program that compares my from-scratch ridge regression using SAG training with the scikit Ridge module shows nearly identical (to 4 decimals) results for the model weights and bias:

Begin SAG ridge regression scratch Python

Loading train (200) and test (40) data

First three train X:
[-0.1660  0.4406 -0.9998 -0.3953 -0.7065]
[ 0.0776 -0.1616  0.3704 -0.5911  0.7562]
[-0.9452  0.3409 -0.1654  0.1174 -0.7192]

First three train y:
0.4840
0.1568
0.8054

==========================

Creating scratch Python ridge model
Done

Training model with SAG
Setting lrn_rate = 0.0010
Setting max_epochs = 200
Setting L2 alpha = 0.5000
Done. Used 127 epochs

Model weights:
[-0.2637  0.0332 -0.0454  0.0355 -0.1139]
Model bias = 0.3620

Evaluating model

Accuracy (within 0.10) train = 0.4650
Accuracy (within 0.10) test = 0.6250

MSE train = 0.00258
MSE test = 0.00195

Predicting for x =
[-0.1660  0.4406 -0.9998 -0.3953 -0.7065]
Predicted y = 0.5322

==========================

Creating scikit Ridge model
Setting L2 alpha = 0.5000
Done

Training scikit Ridge model (using SAG)
Using default training params
Done. Used [36] iterations

Model weights:
[-0.2637  0.0332 -0.0453  0.0355 -0.1139]
Model bias = 0.3619

End demo

Note that for simplicity, I did not scale the training data to mean = 0 and variance = 1, which is more or less required so that all model weights are penalized equally.

I’m not sure what the moral of this blog post is, but it points out the trade-offs between code simplicity and system performance and ease-of-use.


It seems like most people have a passion for something in life. For me, I can’t resist exploring interesting algorithms. For many of my friends, they love games of all types. Here’s an image of the “James Bond Thunderball” board game (1965) from the Milton Bradley company. According to the Internet, this game is quite rare and is highly prized by collectors. I’m pretty sure I would enjoy playing this game, but only once — I don’t think the game is rich enough in strategy to make me want to play it multiple times.

Demo program. Replace

# ridge_regression_sag_scratch.py
# Stochastic Average Gradient training

import numpy as np

# -----------------------------------------------------------

def accuracy(model, data_X, data_y, pct_close):
  n = len(data_X)
  n_correct = 0; n_wrong = 0
  for i in range(n):
    # x = data_X[i].reshape(1,-1)
    x = data_X[i]
    y = data_y[i]
    y_pred = model.predict(x)

    if np.abs(y - y_pred) "lt" np.abs(y * pct_close):
      n_correct += 1
    else: 
      n_wrong += 1
  return n_correct / (n_correct + n_wrong)

def mse(model, data_X, data_y):
  n = len(data_X)
  sum = 0.0
  for i in range(n):
    y = data_y[i]
    pred_y = model.predict(data_X[i])
    diff = y - pred_y
    sum += diff * diff
  return sum /n

# -----------------------------------------------------------

class MyRidge:
  # ridge regression using stochastic average gradient (SAG)
  def __init__(self, seed=0):
    self.weights = None
    self.bias = 0.0
    self.rnd = np.random.RandomState(seed)

  def predict(self, x):
    sum = 0.0
    for j in range(len(x)):
      sum += self.weights[j] * x[j]
    sum += self.bias
    return sum

  def train(self, train_X, train_y, lrn_rate=0.001,
    max_epochs=1000, tol=0.0001, alpha=1.0):
    # using SAG algorithm
    n = len(train_X)
    dim = len(train_X[0])
    self.weights = np.zeros(dim)
    lo = -0.01; hi = 0.01
    for j in range(dim):
      self.weights[j] = (hi - lo) * self.rnd.random() + lo

    # center X and y
    # col_means = np.mean(train_X, axis=0)
    # centered_X = train_X - col_means
    # y_mean = np.mean(train_y)
    # centered_y = train_y - y_mean
    # train_X = centered_X
    # train_y = centered_y

    grad_mem = np.zeros((n, dim))
    tot_grad = np.zeros(dim)
    new_grad = np.zeros(dim)
    indices = np.arange(n)
    alpha_scaled = alpha / n

    prev_weights = np.zeros(dim)

    for epoch in range(max_epochs): # each epoch
      self.rnd.shuffle(indices)
      prev_weights = np.copy(self.weights)
      for i in range(n): # each item
        idx = indices[i]
        x = train_X[idx]
        y = train_y[idx]
        pred_y = self.predict(x)
        err = pred_y - y
        new_grad = err * x
        
        for j in range(dim): # each weight
          tot_grad[j] = tot_grad[j] - grad_mem[idx][j] + \
            new_grad[j]
          grad_mem[idx][j] = new_grad[j]
          self.weights[j] -= lrn_rate * (tot_grad[j] / n)
          self.weights[j] -= lrn_rate * (alpha_scaled * \
            self.weights[j])
        self.bias -= lrn_rate * err

      # check for early stop
      # print("new weights"); print(self.weights)
      # print("prev weights "); print(prev_weights); input()
      max_change_in_wts = np.abs(self.weights - \
        prev_weights).max()
      max_weights = np.abs(self.weights).max()
      # print("max_change_in_wts"); print(max_change_in_wts);
      # print("max_weights"); print(max_weights); input()
      if max_change_in_wts / max_weights "lt" tol:
        return epoch
        # print("Early exit at epoch " + str(epoch))
        # break
      
    return epoch

# -----------------------------------------------------------

print("\nBegin SAG ridge regression scratch Python ")

np.set_printoptions(precision=4, suppress=True,
    floatmode='fixed')

print("\nLoading train (200) and test (40) data ")
train_Xy = np.loadtxt(".\\Data\\synthetic_train_200.txt",
  usecols=[0,1,2,3,4,5], delimiter=",")
train_X = train_Xy[:,[0,1,2,3,4]]
train_y = train_Xy[:,5]

test_Xy = np.loadtxt(".\\Data\\synthetic_test_40.txt",
  usecols=[0,1,2,3,4,5], delimiter=",")
test_X = test_Xy[:,[0,1,2,3,4]]
test_y = test_Xy[:,5]

print("\nFirst three train X: ")
for i in range(3):
  print(train_X[i])
print("\nFirst three train y: ")
for i in range(3):
  print("%0.4f " % train_y[i])

print("\n========================== ")

print("\nCreating scratch Python ridge model ")
model = MyRidge()
print("Done ")

print("\nTraining model with SAG ")
lrn_rate = 0.001
max_epochs = 200
alpha = 0.5
print("Setting lrn_rate = %0.4f " % lrn_rate)
print("Setting max_epochs = " + str(max_epochs))
print("Setting L2 alpha = %0.4f " % alpha)
eps = model.train(train_X, train_y, lrn_rate,
  max_epochs, 0.00001, alpha)
print("Done. Used " + str(eps) + " epochs " )

print("\nModel weights: ")
print(model.weights)
print("Model bias = %0.4f " % model.bias)

print("\nEvaluating model ")
acc_train = accuracy(model, train_X, train_y, 0.10)
acc_test = accuracy(model, test_X, test_y, 0.10)
print("\nAccuracy (within 0.10) train = %0.4f " % \
  acc_train)
print("Accuracy (within 0.10) test = %0.4f " % \
  acc_test)

mse_train = mse(model, train_X, train_y)
mse_test = mse(model, test_X, test_y)
print("\nMSE train = %0.5f " % mse_train)
print("MSE test = %0.5f " % mse_test)

x = train_X[0]
print("\nPredicting for x = ")
print(x)
pred_y = model.predict(x)
print("Predicted y = %0.4f " % pred_y)

print("\n========================== ")

from sklearn.linear_model import Ridge

# Ridge(alpha=1.0, *, fit_intercept=True, copy_X=True,
# max_iter=None, tol=0.0001, solver='auto', positive=False,
# random_state=None)

print("\nCreating scikit Ridge model ")
alpha = 0.5
print("Setting L2 alpha = %0.4f " % alpha)
model = Ridge(alpha=alpha, solver='sag', fit_intercept=True)
print("Done ")

print("\nTraining scikit Ridge model (using SAG) ")
print("Using default training params ")
model.fit(train_X, train_y)
print("Done. Used " + str(model.n_iter_) + " iterations" )

print("\nModel weights: ")
print(model.coef_)
print("Model bias = %0.4f " % model.intercept_)

print("\nEnd demo ")

Training data:

# synthetic_train_200.txt
#
-0.1660,  0.4406, -0.9998, -0.3953, -0.7065,  0.4840
 0.0776, -0.1616,  0.3704, -0.5911,  0.7562,  0.1568
-0.9452,  0.3409, -0.1654,  0.1174, -0.7192,  0.8054
 0.9365, -0.3732,  0.3846,  0.7528,  0.7892,  0.1345
-0.8299, -0.9219, -0.6603,  0.7563, -0.8033,  0.7955
 0.0663,  0.3838, -0.3690,  0.3730,  0.6693,  0.3206
-0.9634,  0.5003,  0.9777,  0.4963, -0.4391,  0.7377
-0.1042,  0.8172, -0.4128, -0.4244, -0.7399,  0.4801
-0.9613,  0.3577, -0.5767, -0.4689, -0.0169,  0.6861
-0.7065,  0.1786,  0.3995, -0.7953, -0.1719,  0.5569
 0.3888, -0.1716, -0.9001,  0.0718,  0.3276,  0.2500
 0.1731,  0.8068, -0.7251, -0.7214,  0.6148,  0.3297
-0.2046, -0.6693,  0.8550, -0.3045,  0.5016,  0.2129
 0.2473,  0.5019, -0.3022, -0.4601,  0.7918,  0.2613
-0.1438,  0.9297,  0.3269,  0.2434, -0.7705,  0.5171
 0.1568, -0.1837, -0.5259,  0.8068,  0.1474,  0.3307
-0.9943,  0.2343, -0.3467,  0.0541,  0.7719,  0.5581
 0.2467, -0.9684,  0.8589,  0.3818,  0.9946,  0.1092
-0.6553, -0.7257,  0.8652,  0.3936, -0.8680,  0.7018
 0.8460,  0.4230, -0.7515, -0.9602, -0.9476,  0.1996
-0.9434, -0.5076,  0.7201,  0.0777,  0.1056,  0.5664
 0.9392,  0.1221, -0.9627,  0.6013, -0.5341,  0.1533
 0.6142, -0.2243,  0.7271,  0.4942,  0.1125,  0.1661
 0.4260,  0.1194, -0.9749, -0.8561,  0.9346,  0.2230
 0.1362, -0.5934, -0.4953,  0.4877, -0.6091,  0.3810
 0.6937, -0.5203, -0.0125,  0.2399,  0.6580,  0.1460
-0.6864, -0.9628, -0.8600, -0.0273,  0.2127,  0.5387
 0.9772,  0.1595, -0.2397,  0.1019,  0.4907,  0.1611
 0.3385, -0.4702, -0.8673, -0.2598,  0.2594,  0.2270
-0.8669, -0.4794,  0.6095, -0.6131,  0.2789,  0.4700
 0.0493,  0.8496, -0.4734, -0.8681,  0.4701,  0.3516
 0.8639, -0.9721, -0.5313,  0.2336,  0.8980,  0.1412
 0.9004,  0.1133,  0.8312,  0.2831, -0.2200,  0.1782
 0.0991,  0.8524,  0.8375, -0.2102,  0.9265,  0.2150
-0.6521, -0.7473, -0.7298,  0.0113, -0.9570,  0.7422
 0.6190, -0.3105,  0.8802,  0.1640,  0.7577,  0.1056
 0.6895,  0.8108, -0.0802,  0.0927,  0.5972,  0.2214
 0.1982, -0.9689,  0.1870, -0.1326,  0.6147,  0.1310
-0.3695,  0.7858,  0.1557, -0.6320,  0.5759,  0.3773
-0.1596,  0.3581,  0.8372, -0.9992,  0.9535,  0.2071
-0.2468,  0.9476,  0.2094,  0.6577,  0.1494,  0.4132
 0.1737,  0.5000,  0.7166,  0.5102,  0.3961,  0.2611
 0.7290, -0.3546,  0.3416, -0.0983, -0.2358,  0.1332
-0.3652,  0.2438, -0.1395,  0.9476,  0.3556,  0.4170
-0.6029, -0.1466, -0.3133,  0.5953,  0.7600,  0.4334
-0.4596, -0.4953,  0.7098,  0.0554,  0.6043,  0.2775
 0.1450,  0.4663,  0.0380,  0.5418,  0.1377,  0.2931
-0.8636, -0.2442, -0.8407,  0.9656, -0.6368,  0.7429
 0.6237,  0.7499,  0.3768,  0.1390, -0.6781,  0.2185
-0.5499,  0.1850, -0.3755,  0.8326,  0.8193,  0.4399
-0.4858, -0.7782, -0.6141, -0.0008,  0.4572,  0.4197
 0.7033, -0.1683,  0.2334, -0.5327, -0.7961,  0.1776
 0.0317, -0.0457, -0.6947,  0.2436,  0.0880,  0.3345
 0.5031, -0.5559,  0.0387,  0.5706, -0.9553,  0.3107
-0.3513,  0.7458,  0.6894,  0.0769,  0.7332,  0.3170
 0.2205,  0.5992, -0.9309,  0.5405,  0.4635,  0.3532
-0.4806, -0.4859,  0.2646, -0.3094,  0.5932,  0.3202
 0.9809, -0.3995, -0.7140,  0.8026,  0.0831,  0.1600
 0.9495,  0.2732,  0.9878,  0.0921,  0.0529,  0.1289
-0.9476, -0.6792,  0.4913, -0.9392, -0.2669,  0.5966
 0.7247,  0.3854,  0.3819, -0.6227, -0.1162,  0.1550
-0.5922, -0.5045, -0.4757,  0.5003, -0.0860,  0.5863
-0.8861,  0.0170, -0.5761,  0.5972, -0.4053,  0.7301
 0.6877, -0.2380,  0.4997,  0.0223,  0.0819,  0.1404
 0.9189,  0.6079, -0.9354,  0.4188, -0.0700,  0.1907
-0.1428, -0.7820,  0.2676,  0.6059,  0.3936,  0.2790
 0.5324, -0.3151,  0.6917, -0.1425,  0.6480,  0.1071
-0.8432, -0.9633, -0.8666, -0.0828, -0.7733,  0.7784
-0.9444,  0.5097, -0.2103,  0.4939, -0.0952,  0.6787
-0.0520,  0.6063, -0.1952,  0.8094, -0.9259,  0.4836
 0.5477, -0.7487,  0.2370, -0.9793,  0.0773,  0.1241
 0.2450,  0.8116,  0.9799,  0.4222,  0.4636,  0.2355
 0.8186, -0.1983, -0.5003, -0.6531, -0.7611,  0.1511
-0.4714,  0.6382, -0.3788,  0.9648, -0.4667,  0.5950
 0.0673, -0.3711,  0.8215, -0.2669, -0.1328,  0.2677
-0.9381,  0.4338,  0.7820, -0.9454,  0.0441,  0.5518
-0.3480,  0.7190,  0.1170,  0.3805, -0.0943,  0.4724
-0.9813,  0.1535, -0.3771,  0.0345,  0.8328,  0.5438
-0.1471, -0.5052, -0.2574,  0.8637,  0.8737,  0.3042
-0.5454, -0.3712, -0.6505,  0.2142, -0.1728,  0.5783
 0.6327, -0.6297,  0.4038, -0.5193,  0.1484,  0.1153
-0.5424,  0.3282, -0.0055,  0.0380, -0.6506,  0.6613
 0.1414,  0.9935,  0.6337,  0.1887,  0.9520,  0.2540
-0.9351, -0.8128, -0.8693, -0.0965, -0.2491,  0.7353
 0.9507, -0.6640,  0.9456,  0.5349,  0.6485,  0.1059
-0.0462, -0.9737, -0.2940, -0.0159,  0.4602,  0.2606
-0.0627, -0.0852, -0.7247, -0.9782,  0.5166,  0.2977
 0.0478,  0.5098, -0.0723, -0.7504, -0.3750,  0.3335
 0.0090,  0.3477,  0.5403, -0.7393, -0.9542,  0.4415
-0.9748,  0.3449,  0.3736, -0.1015,  0.8296,  0.4358
 0.2887, -0.9895, -0.0311,  0.7186,  0.6608,  0.2057
 0.1570, -0.4518,  0.1211,  0.3435, -0.2951,  0.3244
 0.7117, -0.6099,  0.4946, -0.4208,  0.5476,  0.1096
-0.2929, -0.5726,  0.5346, -0.3827,  0.4665,  0.2465
 0.4889, -0.5572, -0.5718, -0.6021, -0.7150,  0.2163
-0.7782,  0.3491,  0.5996, -0.8389, -0.5366,  0.6516
-0.5847,  0.8347,  0.4226,  0.1078, -0.3910,  0.6134
 0.8469,  0.4121, -0.0439, -0.7476,  0.9521,  0.1571
-0.6803, -0.5948, -0.1376, -0.1916, -0.7065,  0.7156
 0.2878,  0.5086, -0.5785,  0.2019,  0.4979,  0.2980
 0.2764,  0.1943, -0.4090,  0.4632,  0.8906,  0.2960
-0.8877,  0.6705, -0.6155, -0.2098, -0.3998,  0.7107
-0.8398,  0.8093, -0.2597,  0.0614, -0.0118,  0.6502
-0.8476,  0.0158, -0.4769, -0.2859, -0.7839,  0.7715
 0.5751, -0.7868,  0.9714, -0.6457,  0.1448,  0.1175
 0.4802, -0.7001,  0.1022, -0.5668,  0.5184,  0.1090
 0.4458, -0.6469,  0.7239, -0.9604,  0.7205,  0.0779
 0.5175,  0.4339,  0.9747, -0.4438, -0.9924,  0.2879
 0.8678,  0.7158,  0.4577,  0.0334,  0.4139,  0.1678
 0.5406,  0.5012,  0.2264, -0.1963,  0.3946,  0.2088
-0.9938,  0.5498,  0.7928, -0.5214, -0.7585,  0.7687
 0.7661,  0.0863, -0.4266, -0.7233, -0.4197,  0.1466
 0.2277, -0.3517, -0.0853, -0.1118,  0.6563,  0.1767
 0.3499, -0.5570, -0.0655, -0.3705,  0.2537,  0.1632
 0.7547, -0.1046,  0.5689, -0.0861,  0.3125,  0.1257
 0.8186,  0.2110,  0.5335,  0.0094, -0.0039,  0.1391
 0.6858, -0.8644,  0.1465,  0.8855,  0.0357,  0.1845
-0.4967,  0.4015,  0.0805,  0.8977,  0.2487,  0.4663
 0.6760, -0.9841,  0.9787, -0.8446, -0.3557,  0.1509
-0.1203, -0.4885,  0.6054, -0.0443, -0.7313,  0.4854
 0.8557,  0.7919, -0.0169,  0.7134, -0.1628,  0.2002
 0.0115, -0.6209,  0.9300, -0.4116, -0.7931,  0.4052
-0.7114, -0.9718,  0.4319,  0.1290,  0.5892,  0.3661
 0.3915,  0.5557, -0.1870,  0.2955, -0.6404,  0.2954
-0.3564, -0.6548, -0.1827, -0.5172, -0.1862,  0.4622
 0.2392, -0.4959,  0.5857, -0.1341, -0.2850,  0.2470
-0.3394,  0.3947, -0.4627,  0.6166, -0.4094,  0.5325
 0.7107,  0.7768, -0.6312,  0.1707,  0.7964,  0.2757
-0.1078,  0.8437, -0.4420,  0.2177,  0.3649,  0.4028
-0.3139,  0.5595, -0.6505, -0.3161, -0.7108,  0.5546
 0.4335,  0.3986,  0.3770, -0.4932,  0.3847,  0.1810
-0.2562, -0.2894, -0.8847,  0.2633,  0.4146,  0.4036
 0.2272,  0.2966, -0.6601, -0.7011,  0.0284,  0.2778
-0.0743, -0.1421, -0.0054, -0.6770, -0.3151,  0.3597
-0.4762,  0.6891,  0.6007, -0.1467,  0.2140,  0.4266
-0.4061,  0.7193,  0.3432,  0.2669, -0.7505,  0.6147
-0.0588,  0.9731,  0.8966,  0.2902, -0.6966,  0.4955
-0.0627, -0.1439,  0.1985,  0.6999,  0.5022,  0.3077
 0.1587,  0.8494, -0.8705,  0.9827, -0.8940,  0.4263
-0.7850,  0.2473, -0.9040, -0.4308, -0.8779,  0.7199
 0.4070,  0.3369, -0.2428, -0.6236,  0.4940,  0.2215
-0.0242,  0.0513, -0.9430,  0.2885, -0.2987,  0.3947
-0.5416, -0.1322, -0.2351, -0.0604,  0.9590,  0.3683
 0.1055,  0.7783, -0.2901, -0.5090,  0.8220,  0.2984
-0.9129,  0.9015,  0.1128, -0.2473,  0.9901,  0.4776
-0.9378,  0.1424, -0.6391,  0.2619,  0.9618,  0.5368
 0.7498, -0.0963,  0.4169,  0.5549, -0.0103,  0.1614
-0.2612, -0.7156,  0.4538, -0.0460, -0.1022,  0.3717
 0.7720,  0.0552, -0.1818, -0.4622, -0.8560,  0.1685
-0.4177,  0.0070,  0.9319, -0.7812,  0.3461,  0.3052
-0.0001,  0.5542, -0.7128, -0.8336, -0.2016,  0.3803
 0.5356, -0.4194, -0.5662, -0.9666, -0.2027,  0.1776
-0.2378,  0.3187, -0.8582, -0.6948, -0.9668,  0.5474
-0.1947, -0.3579,  0.1158,  0.9869,  0.6690,  0.2992
 0.3992,  0.8365, -0.9205, -0.8593, -0.0520,  0.3154
-0.0209,  0.0793,  0.7905, -0.1067,  0.7541,  0.1864
-0.4928, -0.4524, -0.3433,  0.0951, -0.5597,  0.6261
-0.8118,  0.7404, -0.5263, -0.2280,  0.1431,  0.6349
 0.0516, -0.8480,  0.7483,  0.9023,  0.6250,  0.1959
-0.3212,  0.1093,  0.9488, -0.3766,  0.3376,  0.2735
-0.3481,  0.5490, -0.3484,  0.7797,  0.5034,  0.4379
-0.5785, -0.9170, -0.3563, -0.9258,  0.3877,  0.4121
 0.3407, -0.1391,  0.5356,  0.0720, -0.9203,  0.3458
-0.3287, -0.8954,  0.2102,  0.0241,  0.2349,  0.3247
-0.1353,  0.6954, -0.0919, -0.9692,  0.7461,  0.3338
 0.9036, -0.8982, -0.5299, -0.8733, -0.1567,  0.1187
 0.7277, -0.8368, -0.0538, -0.7489,  0.5458,  0.0830
 0.9049,  0.8878,  0.2279,  0.9470, -0.3103,  0.2194
 0.7957, -0.1308, -0.5284,  0.8817,  0.3684,  0.2172
 0.4647, -0.4931,  0.2010,  0.6292, -0.8918,  0.3371
-0.7390,  0.6849,  0.2367,  0.0626, -0.5034,  0.7039
-0.1567, -0.8711,  0.7940, -0.5932,  0.6525,  0.1710
 0.7635, -0.0265,  0.1969,  0.0545,  0.2496,  0.1445
 0.7675,  0.1354, -0.7698, -0.5460,  0.1920,  0.1728
-0.5211, -0.7372, -0.6763,  0.6897,  0.2044,  0.5217
 0.1913,  0.1980,  0.2314, -0.8816,  0.5006,  0.1998
 0.8964,  0.0694, -0.6149,  0.5059, -0.9854,  0.1825
 0.1767,  0.7104,  0.2093,  0.6452,  0.7590,  0.2832
-0.3580, -0.7541,  0.4426, -0.1193, -0.7465,  0.5657
-0.5996,  0.5766, -0.9758, -0.3933, -0.9572,  0.6800
 0.9950,  0.1641, -0.4132,  0.8579,  0.0142,  0.2003
-0.4717, -0.3894, -0.2567, -0.5111,  0.1691,  0.4266
 0.3917, -0.8561,  0.9422,  0.5061,  0.6123,  0.1212
-0.0366, -0.1087,  0.3449, -0.1025,  0.4086,  0.2475
 0.3633,  0.3943,  0.2372, -0.6980,  0.5216,  0.1925
-0.5325, -0.6466, -0.2178, -0.3589,  0.6310,  0.3568
 0.2271,  0.5200, -0.1447, -0.8011, -0.7699,  0.3128
 0.6415,  0.1993,  0.3777, -0.0178, -0.8237,  0.2181
-0.5298, -0.0768, -0.6028, -0.9490,  0.4588,  0.4356
 0.6870, -0.1431,  0.7294,  0.3141,  0.1621,  0.1632
-0.5985,  0.0591,  0.7889, -0.3900,  0.7419,  0.2945
 0.3661,  0.7984, -0.8486,  0.7572, -0.6183,  0.3449
 0.6995,  0.3342, -0.3113, -0.6972,  0.2707,  0.1712
 0.2565,  0.9126,  0.1798, -0.6043, -0.1413,  0.2893
-0.3265,  0.9839, -0.2395,  0.9854,  0.0376,  0.4770
 0.2690, -0.1722,  0.9818,  0.8599, -0.7015,  0.3954
-0.2102, -0.0768,  0.1219,  0.5607, -0.0256,  0.3949
 0.8216, -0.9555,  0.6422, -0.6231,  0.3715,  0.0801
-0.2896,  0.9484, -0.7545, -0.6249,  0.7789,  0.4370
-0.9985, -0.5448, -0.7092, -0.5931,  0.7926,  0.5402

Test data:

# synthetic_test_40.txt
#
 0.7462,  0.4006, -0.0590,  0.6543, -0.0083,  0.1935
 0.8495, -0.2260, -0.0142, -0.4911,  0.7699,  0.1078
-0.2335, -0.4049,  0.4352, -0.6183, -0.7636,  0.5088
 0.1810, -0.5142,  0.2465,  0.2767, -0.3449,  0.3136
-0.8650,  0.7611, -0.0801,  0.5277, -0.4922,  0.7140
-0.2358, -0.7466, -0.5115, -0.8413, -0.3943,  0.4533
 0.4834,  0.2300,  0.3448, -0.9832,  0.3568,  0.1360
-0.6502, -0.6300,  0.6885,  0.9652,  0.8275,  0.3046
-0.3053,  0.5604,  0.0929,  0.6329, -0.0325,  0.4756
-0.7995,  0.0740, -0.2680,  0.2086,  0.9176,  0.4565
-0.2144, -0.2141,  0.5813,  0.2902, -0.2122,  0.4119
-0.7278, -0.0987, -0.3312, -0.5641,  0.8515,  0.4438
 0.3793,  0.1976,  0.4933,  0.0839,  0.4011,  0.1905
-0.8568,  0.9573, -0.5272,  0.3212, -0.8207,  0.7415
-0.5785,  0.0056, -0.7901, -0.2223,  0.0760,  0.5551
 0.0735, -0.2188,  0.3925,  0.3570,  0.3746,  0.2191
 0.1230, -0.2838,  0.2262,  0.8715,  0.1938,  0.2878
 0.4792, -0.9248,  0.5295,  0.0366, -0.9894,  0.3149
-0.4456,  0.0697,  0.5359, -0.8938,  0.0981,  0.3879
 0.8629, -0.8505, -0.4464,  0.8385,  0.5300,  0.1769
 0.1995,  0.6659,  0.7921,  0.9454,  0.9970,  0.2330
-0.0249, -0.3066, -0.2927, -0.4923,  0.8220,  0.2437
 0.4513, -0.9481, -0.0770, -0.4374, -0.9421,  0.2879
-0.3405,  0.5931, -0.3507, -0.3842,  0.8562,  0.3987
 0.9538,  0.0471,  0.9039,  0.7760,  0.0361,  0.1706
-0.0887,  0.2104,  0.9808,  0.5478, -0.3314,  0.4128
-0.8220, -0.6302,  0.0537, -0.1658,  0.6013,  0.4306
-0.4123, -0.2880,  0.9074, -0.0461, -0.4435,  0.5144
 0.0060,  0.2867, -0.7775,  0.5161,  0.7039,  0.3599
-0.7968, -0.5484,  0.9426, -0.4308,  0.8148,  0.2979
 0.7811,  0.8450, -0.6877,  0.7594,  0.2640,  0.2362
-0.6802, -0.1113, -0.8325, -0.6694, -0.6056,  0.6544
 0.3821,  0.1476,  0.7466, -0.5107,  0.2592,  0.1648
 0.7265,  0.9683, -0.9803, -0.4943, -0.5523,  0.2454
-0.9049, -0.9797, -0.0196, -0.9090, -0.4433,  0.6447
-0.4607,  0.1811, -0.2389,  0.4050, -0.0078,  0.5229
 0.2664, -0.2932, -0.4259, -0.7336,  0.8742,  0.1834
-0.4507,  0.1029, -0.6294, -0.1158, -0.6294,  0.6081
 0.8948, -0.0124,  0.9278,  0.2899, -0.0314,  0.1534
-0.1323, -0.8813, -0.0146, -0.0697,  0.6135,  0.2386
Posted in Machine Learning, Scikit | Leave a comment

Refactoring My Trimmed Kernel Ridge Regression to Approximate Support Vector Regression Using C#

Posted on June 15, 2026 by jamesdmccaffrey

Bottom line: I refactored my C# implementation of trimmed kernel ridge regression to approximate support vector regression. The refactoring works fine, but the code is so long, there are multiple ways to organize it, and so I might have to refactor at some point in the future.

As is often the case with interesting topics, explaining what the problem is, takes longer than explaining the solution. So bear with me.

The goal of a machine learning regression problem is to predict a single numeric value. For example, a bank might want to predict a maximum loan amount based on applicant age, sex, annual income, debt, and so on.

Common regression techniques include linear regression, quadratic regression, nearest neighbors regression, kernel ridge regression, Gaussian process regression, kernel support vector regression, neural network regression, random forest regression, and gradient boost regression. Each technique has dozens of variations, and each technique has pros and cons.

Two closely related machine learning regression techniques are kernel ridge regression (KRR) and support vector regression (SVR). Both techniques use a kernel function (usually the radial basis function, RBF) to compare two data items for similarity. Both techniques must store training data in order to make predictions, but KRR must store all training items, while SVR eliminates some of the items during training, leaving just the “support vectors” that need to be stored.

On the other hand, the loss function used by SVR is not differentiable, so SVR cannot be trained using stochastic gradient descent, and therefore, SVR does not scale well to very large datasets.

Note: In addition to the kernel support vector regression technique discussed in this blog post, there is a linear support vector regression technique which is absolutely useless in practice.

KRR often gives slightly better prediction accuracy than SVR, is easier to train than SVR, can handle very large datasets via SGD training, and is much, much easier to implement than SVR.

One morning before work, I got the idea of combining KRR and SVR. Briefly, I train a KRR model as normal using all data. Then I identify training data items that are predicted “too well” and remove them, leaving just pseudo support vectors. The somewhat counterintuitive SVR idea is that items that are predicted too well don’t help model accuracy very much. Then I retrain a new KRR model using only the reduced training data. This gives a trimmed/sparse KRR model that approximates an SVR model.

In my mind, this technique gives the advantages of KRR (ability to handle very large datasets) and the advantages of SVR (fewer stored model items and weights than KRR). I first implemented the idea using the scikit KernelRidge module and the idea worked well. Then I implemented the idea using C#, just to see if I could — using C# I had to implent a lot of code from scratch that is built-in to Python and scikit. The C# code worked but it was clunky so — the point of this blog post — I refactored the clunky C# code to make it more rational.

Let me point out that the idea of trimming a KRR model to approximate an SVR model is fairly obvious, and so I’m reasonably sure other people have used the idea, but I could find no examples or discussion on the Internet.

The output of the new, refactored C# demo is:

Begin kernel ridge regression trim 
approximation to SVR using C#

Loading train (200) and test (40) data
Done

First three X predictors:
  -0.1660   0.4406  -0.9998  -0.3953  -0.7065
   0.0776  -0.1616   0.3704  -0.5911   0.7562
  -0.9452   0.3409  -0.1654   0.1174  -0.7192

First three target y:
  0.4840
  0.1568
  0.8054

Creating KRR model
Setting RBF gamma = 0.30
Done

Setting lrnRate = 0.0500
Setting maxEpochs = 5000
Setting alpha decay =  0.00001
Setting auto-exit tol = 0.0010
Setting trim epsilon = 0.0035

Training model
epoch =      0 MSE = 0.0181 acc = 0.1700
Auto-exit at epoch 723
Auto-exit at epoch 707
Done

Final model weights:
 -0.7406  -0.4260  -0.3825 . . .
. . .      0.3299   0.8800

Evaluating trimmed KRR model
Number model item/weights = 109

Train acc = 0.9817
Test acc = 0.9500

Train MSE = 0.0001
Test MSE = 0.0003

End KRR trim/prune demo

The demo data is synthetic demo. There are 200 training items and 40 test items. The demo data was generated by a neural network with random weights and biases. The data is normalized, which is strogly recommended for KRR and SVR because items are compared mathematically and you don’t want a column with huge magnitudes to dominate.

Both KRR and SVR can handle non-numeric data, but that’s another story.

The key calling statements are:

KRR krr = new KRR(gamma);

double lrnRate = 0.05;
int maxEpochs = 5000;
double alpha = 1.0e-5;  // wt decay regularization
double exitTol = 0.001;  // auto-exit
double epsilon = 0.0035; // trimming
// larger epsilon: fewer wts retained
// smaller epsilon: more wts retained
krr.Train(trainX, trainY, lrnRate, maxEpochs, alpha,
  exitTol, epsilon);

The large number of parameters is a significant weakness of KRR (and SVR too).

After training on 200 data items, those items that were predicted with a small error of less than 0.0035 were removed. There were 91 such well-predicted items, leaving a reduced training dataset of 109 items. These would be called the support vectors in SVR. After training a new model on the reduced dataset, the sparse model scored 98.17% accuracy on the reduced training data (107 out of 109 correct) and 95.00% accuracy on the test data (38 out of 40 correct). Nice!

To validate my C# trimmed KRR approximation to SVR, I ran the data through the scikit SVR module. A grid-search optimized scikit SVR model created 90 support vectors (better than the trimmed KRR 109 items), with 96.50% accuracy on the training data (not quite as good as the C# trimmed KRR model), and 95.00% accuracy on the test data (identical to the sparse KRR model). In short, the C# trimmed KRR model and the SVR model are very similar. The large number of hyperparameters involved with KRR and SVR makes exact comparison impossible in practice.

An interesting experiment.


Like most people I know, I get obsessed with all kinds of strange things. I have been obsessed with machine learning for many years.

For reason(s) unknown to me, I am mildly obsessed with science fiction movies that feature a creature that makes noises that are translated to “chittering” in the closed caption text. The are an amazing number of chittering instances in science fiction movies, that you’ll only notice if you’re looking for them.

Left: In “Tarantua” (1955), a scientist experiments with making animals grow and shrink. What could possibly go wrong? The giant tarantula is eventually disposed of via bombs and missiles from an F-80 jet that is piloted by actor Clint Eastwood.

Right: I avoided watching “Predator: Badlands” (2025) for a long time because all of the previous Predator movies were awful (except for the first in 1987). My good friend Ken L convinced me to watch “Badlands” in his fancy home theater. It was really good, and on top of that, it was a bonanza of “chittering”. The very first opening scene had a small chittering creature and then on the Death Plant, basically every plant and animal was a chitterer.

The (refacored) C# trimmed KRR demo program. Replace “lt” (less than), “gt”, “lte”, “gte” with Boolean operator symbols. (My blog editor chokes on symbols).

using System;
using System.IO;
using System.Collections.Generic;

namespace KernelRidgeRegressionTrim
{
  internal class KernelRidgeRegressionProgram
  {
    static void Main(string[] args)
    {
      Console.WriteLine("\nBegin kernel ridge " +
        "regression trim approximation to SVR using C# ");

      Console.WriteLine("\nLoading train (200) and" +
        " test (40) data ");
      string trainFile =
        "..\\..\\..\\Data\\synthetic_train_200.txt";
      double[][] trainX =
        Utils.MatLoad(trainFile,
        new int[] { 0, 1, 2, 3, 4 }, ',', "#");
      double[] trainY =
        Utils.MatToVec(Utils.MatLoad(trainFile,
        new int[] { 5 }, ',', "#"));

      string testFile =
        "..\\..\\..\\Data\\synthetic_test_40.txt";
      double[][] testX =
        Utils.MatLoad(testFile,
        new int[] { 0, 1, 2, 3, 4 }, ',', "#");
      double[] testY =
        Utils.MatToVec(Utils.MatLoad(testFile,
        new int[] { 5 }, ',', "#"));
      Console.WriteLine("Done ");

      Console.WriteLine("\nFirst three X predictors: ");
      for (int i = 0; i "lt" 3; ++i)
        Utils.VecShow(trainX[i], 4, 9);
      Console.WriteLine("\nFirst three target y: ");
      for (int i = 0; i "lt" 3; ++i)
        Console.WriteLine(trainY[i].ToString("F4").
          PadLeft(8));

      Console.WriteLine("\nCreating KRR model ");
      double gamma = 0.3;    // RBF param
      Console.WriteLine("Setting RBF gamma = " +
        gamma.ToString("F2"));
      KRR krr = new KRR(gamma);
      Console.WriteLine("Done ");

      double lrnRate = 0.05;
      int maxEpochs = 5000;
      double alpha = 1.0e-5;  // wt decay regularization
      double exitTol = 0.001;  // auto-exit
      double epsilon = 0.0035; // trimming
      // larger epsilon: more wts retained
      // smaller epsilon: fewer wts retained
      Console.WriteLine("\nSetting lrnRate = " +
        lrnRate.ToString("F4"));
      Console.WriteLine("Setting maxEpochs = " + maxEpochs);
      Console.WriteLine("Setting alpha decay =  " +
        alpha.ToString("F5"));
      Console.WriteLine("Setting auto-exit tol = " +
        exitTol.ToString("F4"));
      Console.WriteLine("Setting trim epsilon = " +
        epsilon.ToString("F4"));
      Console.WriteLine("\nTraining model ");
      krr.Train(trainX, trainY, lrnRate, maxEpochs, alpha,
        exitTol, epsilon);
      Console.WriteLine("Done ");
      Console.WriteLine("\nFinal model weights: ");
      Utils.VecShow(krr.wts, 4, 9);

      Console.WriteLine("\nEvaluating trimmed KRR model ");
      Console.WriteLine("Number model item/weights = " +
        krr.wts.Length);
      double trainAcc = 
        krr.Accuracy(krr.trainX, krr.trainY, 0.10);
      double testAcc = krr.Accuracy(testX, testY, 0.10);

      Console.WriteLine("\nTrain acc = " +
        trainAcc.ToString("F4"));
      Console.WriteLine("Test acc = " +
        testAcc.ToString("F4"));

      double trainMSE = krr.MSE(krr.trainX, krr.trainY);
      double testMSE = krr.MSE(testX, testY);

      Console.WriteLine("\nTrain MSE = " +
        trainMSE.ToString("F4"));
      Console.WriteLine("Test MSE = " +
        testMSE.ToString("F4"));

      Console.WriteLine("\nEnd KRR trim/prune demo ");
      Console.ReadLine();
    } // Main()

  } // class Program

  /*
  Creating scikit SVR model
  Setting gamma = 0.1000
  Setting C = 10.0000
  Setting epsilon = 0.0100

  Number support vectors = 90
  Accuracy (within 0.10) train = 0.9650
  Accuracy (within 0.10) test = 0.9500
  MSE train = 0.0001
  MSE test = 0.0002 
  */

  // ========================================================

  public class KRR
  {
    public double gamma;  // for RBF kernel
    public double[][] trainX;  // need for prediction
    public double[] trainY; // not used this version
    public double[] wts;  // one per trainX item
    public Random rnd; // for SGD

    // ------------------------------------------------------

    public KRR(double gamma, int seed = 1)
    {
      this.gamma = gamma;
      this.trainX = new double[0][];
      this.trainY = new double[0];
      this.wts = new double[0];
      this.rnd = new Random(seed);  // shuffle train order
    } // ctor

    // ------------------------------------------------------

    private void Shuffle(int[] indices)
    {
      // Fisher-Yates for SGD
      for (int i = 0; i "lt" indices.Length; ++i)
      {
        int ri = this.rnd.Next(i, indices.Length);
        int tmp = indices[i];
        indices[i] = indices[ri];
        indices[ri] = tmp;
      }
    } // Shuffle

    // ------------------------------------------------------

    public void Train(double[][] trainX, double[] trainY,
     double lrnRate, int maxEpochs, double alpha,
     double exitTol, double epsilon)
    {
      int N = trainX.Length; // not used this version
      int freq = maxEpochs / 5;  // when to show progress

      // 0. store trainX -- needed by Predict()
      // copy by value for safety. change to copy
      // by ref if short of memory.
      this.trainX = new double[trainX.Length][];
      for (int i = 0; i "lt" trainX.Length; ++i)
        this.trainX[i] = new double[trainX[0].Length];
      for (int i = 0; i "lt" trainX.Length; ++i)
        for (int j = 0; j "lt" trainX[0].Length; ++j)
          this.trainX[i][j] = trainX[i][j];

      this.trainY = new double[trainY.Length];
      for (int i = 0; i "lt" this.trainY.Length; ++i)
        this.trainY[i] = trainY[i];

      // 1. init weights
      this.wts = new double[trainX.Length];
      double lo = -0.10; double hi = 0.10;
      for (int i = 0; i "lt" this.wts.Length; ++i)
        this.wts[i] = (hi - lo) *
          this.rnd.NextDouble() + lo;

      // 2. set up indices for shuffling
      int[] indices = new int[trainX.Length];
      for (int i = 0; i "lt" indices.Length; ++i)
        indices[i] = i;

      // 3. set up prev weights for auto-exit
      double[] prevWeights = new double[trainX.Length];
      for (int j = 0; j "lt" trainX.Length; ++j)
        prevWeights[j] = this.wts[j];

      // 4. SGD prelim train
      for (int epoch = 0; epoch "lt" maxEpochs; ++epoch)
      {
        Shuffle(indices);
        for (int i = 0; i "lt" trainX.Length; ++i)
        {
          int idx = indices[i];
          double[] x = trainX[idx];
          double predY = this.Predict(x);
          double actualY = trainY[idx];

          // update wt assoc with x
          this.wts[idx] -= lrnRate * (predY - actualY);
        } // each item

        if (epoch % freq == 0)
        {
          double rmse = this.MSE(trainX, trainY);
          double acc = this.Accuracy(trainX, trainY, 0.10);
          string s1 = "epoch = " +
            epoch.ToString().PadLeft(6);
          string s2 = " MSE = " +
            rmse.ToString("F4");
          string s3 = " acc = " + acc.ToString("F4");
          Console.WriteLine(s1 + s2 + s3);
        }
                
        // if max_change_in_wts / max_weights "lt" tol
        int numWts = this.wts.Length;
        double[] weightDeltas = new double[numWts];
        for (int j = 0; j "lt" numWts; ++j)
          weightDeltas[j] = 
            Math.Abs(prevWeights[j] - this.wts[j]);
        double maxChange = weightDeltas[0];
        for (int j = 0; j "lt" numWts; ++j)
          if (weightDeltas[j] "gt" maxChange)
            maxChange = weightDeltas[j];

        double maxWeightMag = Math.Abs(this.wts[0]);
        for (int j = 0; j "lt" numWts; ++j)
          if (Math.Abs(this.wts[j]) "gt" maxWeightMag)
            maxWeightMag = Math.Abs(this.wts[j]);

        if (maxWeightMag != 0.0 &&
          (maxChange / maxWeightMag) "lt" exitTol)
        {
          Console.WriteLine("Auto-exit at epoch " + epoch);
          break;
        }

        // an auto-exit didn't happen
        for (int j = 0; j "lt" numWts; ++j)
          prevWeights[j] = this.wts[j];
      
      } // each epoch

      // 5. apply one final wt decay regularization
      // as opposed to small decay after every epoch
      for (int j = 0; j "lt" this.wts.Length; ++j)
        this.wts[j] *= (1.0 - alpha);
      
      // trimming phase

      // 6. find idxs of items that are predicted too well
      // 0 = not supp vec == predicted too well.
      // 1 = supp vec == predicts only OK.
      int[] isSupportVec = new int[trainX.Length]; 
      for (int i = 0; i "lt" trainX.Length; ++i)
      {
        if (Math.Abs(this.Predict(trainX[i]) - 
          trainY[i]) "gt" epsilon)
          isSupportVec[i] = 1; // just OK prediction
      }
      
      // 7. create a new, smaller training dataset
      List"lt"double[]"gt" tmpTrainX = 
        new List"lt"double[]"gt"();
      List"lt"double"gt" tmpTrainY = 
        new List"lt"double"gt"();
      for (int i = 0; i "lt" trainX.Length; ++i)
      {
        if (isSupportVec[i] == 1)
        {
          tmpTrainX.Add(trainX[i]);
          tmpTrainY.Add(trainY[i]);
        }
      }
      double[][] newTrainX = tmpTrainX.ToArray();
      double[] newTrainY = tmpTrainY.ToArray();

      // 8. re-train using new small dataset
      // store trainX -- needed by Predict()
      this.trainX = newTrainX;  // by ref -- could copy
      this.trainY = newTrainY;  // not used this version

      // re-init new, smaller set of weights
      this.wts = new double[newTrainX.Length];
      for (int i = 0; i "lt" this.wts.Length; ++i)
        this.wts[i] = (hi - lo) *
          this.rnd.NextDouble() + lo;

      // set up new, small indices for shuffling
      int[] newIndices = new int[newTrainX.Length];
      for (int i = 0; i "lt" newIndices.Length; ++i)
        newIndices[i] = i;

      // set up new prev weights for auto-exit
      double[] newPrevWeights = new double[newTrainX.Length];
      for (int j = 0; j "lt" newTrainX.Length; ++j)
        newPrevWeights[j] = this.wts[j];

      // SGD final train
      for (int epoch = 0; epoch "lt" maxEpochs; ++epoch)
      {
        Shuffle(newIndices);
        for (int i = 0; i "lt" newTrainX.Length; ++i)
        {
          int idx = newIndices[i];
          double[] x = newTrainX[idx];
          double predY = this.Predict(x);
          double actualY = newTrainY[idx];

          // update wt assoc with x
          this.wts[idx] -= lrnRate * (predY - actualY);
        } // each item

        // show progress on second SGD
        //if (epoch % freq == 0)
        //{
        //  double rmse = this.MSE(trainX, trainY);
        //  double acc = this.Accuracy(trainX, trainY, 0.10);
        //  string s1 = "epoch = " +
        //    epoch.ToString().PadLeft(6);
        //  string s2 = " MSE = " +
        //    rmse.ToString("F4");
        //  string s3 = " acc = " + acc.ToString("F4");
        //  Console.WriteLine(s1 + s2 + s3);
        //}
                
        // if max_change_in_wts / max_weights "lt" tol
        int numWts = this.wts.Length;
        double[] weightDeltas = new double[numWts];
        for (int j = 0; j "lt" numWts; ++j)
          weightDeltas[j] = 
            Math.Abs(newPrevWeights[j] - this.wts[j]);
        double maxChange = weightDeltas[0];
        for (int j = 0; j "lt" numWts; ++j)
          if (weightDeltas[j] "gt" maxChange)
            maxChange = weightDeltas[j];

        double maxWeightMag = Math.Abs(this.wts[0]);
        for (int j = 0; j "lt" numWts; ++j)
          if (Math.Abs(this.wts[j]) "gt" maxWeightMag)
            maxWeightMag = Math.Abs(this.wts[j]);

        if (maxWeightMag != 0.0 &&
          (maxChange / maxWeightMag) "lt" exitTol)
        {
          Console.WriteLine("Auto-exit at epoch " + epoch);
          break;
        }

        // an auto-exit didn't happen
        for (int j = 0; j "lt" numWts; ++j)
          newPrevWeights[j] = this.wts[j];
      
      } // each epoch

      // apply one final wt decay regularization
      // as opposed to small decay after every epoch
      for (int j = 0; j "lt" this.wts.Length; ++j)
        this.wts[j] *= (1.0 - alpha);

    } // Train()

    // ------------------------------------------------------

    private double Rbf(double[] v1, double[] v2)
    {
      // the gamma version aot len_scale version
      int dim = v1.Length;
      double sum = 0.0;
      for (int i = 0; i "lt" dim; ++i)
      {
        sum += (v1[i] - v2[i]) * (v1[i] - v2[i]);
      }
      return Math.Exp(-1 * this.gamma * sum);
    }

    // ------------------------------------------------------

    public double Predict(double[] x)
    {
      int N = this.trainX.Length;
      double sum = 0.0;
      for (int i = 0; i "lt" N; ++i)
      {
        double[] xx = this.trainX[i];
        double k = this.Rbf(x, xx);
        sum += this.wts[i] * k;
      }
      return sum;
    }

    // ------------------------------------------------------

    public double Accuracy(double[][] dataX,
      double[] dataY, double pctClose)
    {
      int numCorrect = 0; int numWrong = 0;
      int n = dataX.Length;
      for (int i = 0; i "lt" n; ++i)
      {
        double[] x = dataX[i];
        double actualY = dataY[i];
        double predY = this.Predict(x);
        if (Math.Abs(actualY - predY) "lt"
          Math.Abs(actualY * pctClose))
          ++numCorrect;
        else
          ++numWrong;
      }
      return (numCorrect * 1.0) / n;
    }

    // ------------------------------------------------------

    public double MSE(double[][] dataX, double[] dataY)
    {
      double sum = 0.0;
      int n = dataX.Length;
      for (int i = 0; i "lt" n; ++i)
      {
        double[] x = dataX[i];
        double actualY = dataY[i];
        double predY = this.Predict(x);
        sum += (actualY - predY) * (actualY - predY);
      }
      return sum / n;
    }
  } // class KRR

  // ========================================================

  public class Utils
  {
    // ------------------------------------------------------

    public static double[][] MatLoad(string fn,
      int[] usecols, char sep, string comment)
    {
      List"lt"double[]"gt" result = 
        new List"lt"double[]"gt"();
      string line = "";
      FileStream ifs = new FileStream(fn, FileMode.Open);
      StreamReader sr = new StreamReader(ifs);
      while ((line = sr.ReadLine()) != null)
      {
        if (line.StartsWith(comment) == true)
          continue;
        string[] tokens = line.Split(sep);
        List"lt"double"gt" lst = new List"lt"double"gt"();
        for (int j = 0; j "lt" usecols.Length; ++j)
          lst.Add(double.Parse(tokens[usecols[j]]));
        double[] row = lst.ToArray();
        result.Add(row);
      }
      sr.Close(); ifs.Close();
      return result.ToArray();
    }

    // ------------------------------------------------------

    public static double[] MatToVec(double[][] mat)
    {
      int nRows = mat.Length;
      int nCols = mat[0].Length;
      double[] result = new double[nRows * nCols];
      int k = 0;
      for (int i = 0; i "lt" nRows; ++i)
        for (int j = 0; j "lt" nCols; ++j)
          result[k++] = mat[i][j];
      return result;
    }

    // ------------------------------------------------------

    public static void MatShow(double[][] m, int dec,
      int wid)
    {
      int nRows = m.Length; int nCols = m[0].Length;
      double small = 1.0 / Math.Pow(10, dec);
      for (int i = 0; i "lt" nRows; ++i)
      {
        for (int j = 0; j "lt" nCols; ++j)
        {
          double v = m[i][j];
          if (Math.Abs(v) "lt" small) v = 0.0;
          Console.Write(v.ToString("F" + dec).
            PadLeft(wid));
        }
        Console.WriteLine("");
      }
    }

    // ------------------------------------------------------

    public static void VecShow(double[] vec, int dec,
      int wid)
    {
      for (int i = 0; i "lt" vec.Length; ++i)
        Console.Write(vec[i].ToString("F" + dec).
          PadLeft(wid));
      Console.WriteLine("");
    }

    // ------------------------------------------------------

  } // class Utils

  // ========================================================

} // ns

Training data:

# synthetic_train_200.txt
#
-0.1660,  0.4406, -0.9998, -0.3953, -0.7065,  0.4840
 0.0776, -0.1616,  0.3704, -0.5911,  0.7562,  0.1568
-0.9452,  0.3409, -0.1654,  0.1174, -0.7192,  0.8054
 0.9365, -0.3732,  0.3846,  0.7528,  0.7892,  0.1345
-0.8299, -0.9219, -0.6603,  0.7563, -0.8033,  0.7955
 0.0663,  0.3838, -0.3690,  0.3730,  0.6693,  0.3206
-0.9634,  0.5003,  0.9777,  0.4963, -0.4391,  0.7377
-0.1042,  0.8172, -0.4128, -0.4244, -0.7399,  0.4801
-0.9613,  0.3577, -0.5767, -0.4689, -0.0169,  0.6861
-0.7065,  0.1786,  0.3995, -0.7953, -0.1719,  0.5569
 0.3888, -0.1716, -0.9001,  0.0718,  0.3276,  0.2500
 0.1731,  0.8068, -0.7251, -0.7214,  0.6148,  0.3297
-0.2046, -0.6693,  0.8550, -0.3045,  0.5016,  0.2129
 0.2473,  0.5019, -0.3022, -0.4601,  0.7918,  0.2613
-0.1438,  0.9297,  0.3269,  0.2434, -0.7705,  0.5171
 0.1568, -0.1837, -0.5259,  0.8068,  0.1474,  0.3307
-0.9943,  0.2343, -0.3467,  0.0541,  0.7719,  0.5581
 0.2467, -0.9684,  0.8589,  0.3818,  0.9946,  0.1092
-0.6553, -0.7257,  0.8652,  0.3936, -0.8680,  0.7018
 0.8460,  0.4230, -0.7515, -0.9602, -0.9476,  0.1996
-0.9434, -0.5076,  0.7201,  0.0777,  0.1056,  0.5664
 0.9392,  0.1221, -0.9627,  0.6013, -0.5341,  0.1533
 0.6142, -0.2243,  0.7271,  0.4942,  0.1125,  0.1661
 0.4260,  0.1194, -0.9749, -0.8561,  0.9346,  0.2230
 0.1362, -0.5934, -0.4953,  0.4877, -0.6091,  0.3810
 0.6937, -0.5203, -0.0125,  0.2399,  0.6580,  0.1460
-0.6864, -0.9628, -0.8600, -0.0273,  0.2127,  0.5387
 0.9772,  0.1595, -0.2397,  0.1019,  0.4907,  0.1611
 0.3385, -0.4702, -0.8673, -0.2598,  0.2594,  0.2270
-0.8669, -0.4794,  0.6095, -0.6131,  0.2789,  0.4700
 0.0493,  0.8496, -0.4734, -0.8681,  0.4701,  0.3516
 0.8639, -0.9721, -0.5313,  0.2336,  0.8980,  0.1412
 0.9004,  0.1133,  0.8312,  0.2831, -0.2200,  0.1782
 0.0991,  0.8524,  0.8375, -0.2102,  0.9265,  0.2150
-0.6521, -0.7473, -0.7298,  0.0113, -0.9570,  0.7422
 0.6190, -0.3105,  0.8802,  0.1640,  0.7577,  0.1056
 0.6895,  0.8108, -0.0802,  0.0927,  0.5972,  0.2214
 0.1982, -0.9689,  0.1870, -0.1326,  0.6147,  0.1310
-0.3695,  0.7858,  0.1557, -0.6320,  0.5759,  0.3773
-0.1596,  0.3581,  0.8372, -0.9992,  0.9535,  0.2071
-0.2468,  0.9476,  0.2094,  0.6577,  0.1494,  0.4132
 0.1737,  0.5000,  0.7166,  0.5102,  0.3961,  0.2611
 0.7290, -0.3546,  0.3416, -0.0983, -0.2358,  0.1332
-0.3652,  0.2438, -0.1395,  0.9476,  0.3556,  0.4170
-0.6029, -0.1466, -0.3133,  0.5953,  0.7600,  0.4334
-0.4596, -0.4953,  0.7098,  0.0554,  0.6043,  0.2775
 0.1450,  0.4663,  0.0380,  0.5418,  0.1377,  0.2931
-0.8636, -0.2442, -0.8407,  0.9656, -0.6368,  0.7429
 0.6237,  0.7499,  0.3768,  0.1390, -0.6781,  0.2185
-0.5499,  0.1850, -0.3755,  0.8326,  0.8193,  0.4399
-0.4858, -0.7782, -0.6141, -0.0008,  0.4572,  0.4197
 0.7033, -0.1683,  0.2334, -0.5327, -0.7961,  0.1776
 0.0317, -0.0457, -0.6947,  0.2436,  0.0880,  0.3345
 0.5031, -0.5559,  0.0387,  0.5706, -0.9553,  0.3107
-0.3513,  0.7458,  0.6894,  0.0769,  0.7332,  0.3170
 0.2205,  0.5992, -0.9309,  0.5405,  0.4635,  0.3532
-0.4806, -0.4859,  0.2646, -0.3094,  0.5932,  0.3202
 0.9809, -0.3995, -0.7140,  0.8026,  0.0831,  0.1600
 0.9495,  0.2732,  0.9878,  0.0921,  0.0529,  0.1289
-0.9476, -0.6792,  0.4913, -0.9392, -0.2669,  0.5966
 0.7247,  0.3854,  0.3819, -0.6227, -0.1162,  0.1550
-0.5922, -0.5045, -0.4757,  0.5003, -0.0860,  0.5863
-0.8861,  0.0170, -0.5761,  0.5972, -0.4053,  0.7301
 0.6877, -0.2380,  0.4997,  0.0223,  0.0819,  0.1404
 0.9189,  0.6079, -0.9354,  0.4188, -0.0700,  0.1907
-0.1428, -0.7820,  0.2676,  0.6059,  0.3936,  0.2790
 0.5324, -0.3151,  0.6917, -0.1425,  0.6480,  0.1071
-0.8432, -0.9633, -0.8666, -0.0828, -0.7733,  0.7784
-0.9444,  0.5097, -0.2103,  0.4939, -0.0952,  0.6787
-0.0520,  0.6063, -0.1952,  0.8094, -0.9259,  0.4836
 0.5477, -0.7487,  0.2370, -0.9793,  0.0773,  0.1241
 0.2450,  0.8116,  0.9799,  0.4222,  0.4636,  0.2355
 0.8186, -0.1983, -0.5003, -0.6531, -0.7611,  0.1511
-0.4714,  0.6382, -0.3788,  0.9648, -0.4667,  0.5950
 0.0673, -0.3711,  0.8215, -0.2669, -0.1328,  0.2677
-0.9381,  0.4338,  0.7820, -0.9454,  0.0441,  0.5518
-0.3480,  0.7190,  0.1170,  0.3805, -0.0943,  0.4724
-0.9813,  0.1535, -0.3771,  0.0345,  0.8328,  0.5438
-0.1471, -0.5052, -0.2574,  0.8637,  0.8737,  0.3042
-0.5454, -0.3712, -0.6505,  0.2142, -0.1728,  0.5783
 0.6327, -0.6297,  0.4038, -0.5193,  0.1484,  0.1153
-0.5424,  0.3282, -0.0055,  0.0380, -0.6506,  0.6613
 0.1414,  0.9935,  0.6337,  0.1887,  0.9520,  0.2540
-0.9351, -0.8128, -0.8693, -0.0965, -0.2491,  0.7353
 0.9507, -0.6640,  0.9456,  0.5349,  0.6485,  0.1059
-0.0462, -0.9737, -0.2940, -0.0159,  0.4602,  0.2606
-0.0627, -0.0852, -0.7247, -0.9782,  0.5166,  0.2977
 0.0478,  0.5098, -0.0723, -0.7504, -0.3750,  0.3335
 0.0090,  0.3477,  0.5403, -0.7393, -0.9542,  0.4415
-0.9748,  0.3449,  0.3736, -0.1015,  0.8296,  0.4358
 0.2887, -0.9895, -0.0311,  0.7186,  0.6608,  0.2057
 0.1570, -0.4518,  0.1211,  0.3435, -0.2951,  0.3244
 0.7117, -0.6099,  0.4946, -0.4208,  0.5476,  0.1096
-0.2929, -0.5726,  0.5346, -0.3827,  0.4665,  0.2465
 0.4889, -0.5572, -0.5718, -0.6021, -0.7150,  0.2163
-0.7782,  0.3491,  0.5996, -0.8389, -0.5366,  0.6516
-0.5847,  0.8347,  0.4226,  0.1078, -0.3910,  0.6134
 0.8469,  0.4121, -0.0439, -0.7476,  0.9521,  0.1571
-0.6803, -0.5948, -0.1376, -0.1916, -0.7065,  0.7156
 0.2878,  0.5086, -0.5785,  0.2019,  0.4979,  0.2980
 0.2764,  0.1943, -0.4090,  0.4632,  0.8906,  0.2960
-0.8877,  0.6705, -0.6155, -0.2098, -0.3998,  0.7107
-0.8398,  0.8093, -0.2597,  0.0614, -0.0118,  0.6502
-0.8476,  0.0158, -0.4769, -0.2859, -0.7839,  0.7715
 0.5751, -0.7868,  0.9714, -0.6457,  0.1448,  0.1175
 0.4802, -0.7001,  0.1022, -0.5668,  0.5184,  0.1090
 0.4458, -0.6469,  0.7239, -0.9604,  0.7205,  0.0779
 0.5175,  0.4339,  0.9747, -0.4438, -0.9924,  0.2879
 0.8678,  0.7158,  0.4577,  0.0334,  0.4139,  0.1678
 0.5406,  0.5012,  0.2264, -0.1963,  0.3946,  0.2088
-0.9938,  0.5498,  0.7928, -0.5214, -0.7585,  0.7687
 0.7661,  0.0863, -0.4266, -0.7233, -0.4197,  0.1466
 0.2277, -0.3517, -0.0853, -0.1118,  0.6563,  0.1767
 0.3499, -0.5570, -0.0655, -0.3705,  0.2537,  0.1632
 0.7547, -0.1046,  0.5689, -0.0861,  0.3125,  0.1257
 0.8186,  0.2110,  0.5335,  0.0094, -0.0039,  0.1391
 0.6858, -0.8644,  0.1465,  0.8855,  0.0357,  0.1845
-0.4967,  0.4015,  0.0805,  0.8977,  0.2487,  0.4663
 0.6760, -0.9841,  0.9787, -0.8446, -0.3557,  0.1509
-0.1203, -0.4885,  0.6054, -0.0443, -0.7313,  0.4854
 0.8557,  0.7919, -0.0169,  0.7134, -0.1628,  0.2002
 0.0115, -0.6209,  0.9300, -0.4116, -0.7931,  0.4052
-0.7114, -0.9718,  0.4319,  0.1290,  0.5892,  0.3661
 0.3915,  0.5557, -0.1870,  0.2955, -0.6404,  0.2954
-0.3564, -0.6548, -0.1827, -0.5172, -0.1862,  0.4622
 0.2392, -0.4959,  0.5857, -0.1341, -0.2850,  0.2470
-0.3394,  0.3947, -0.4627,  0.6166, -0.4094,  0.5325
 0.7107,  0.7768, -0.6312,  0.1707,  0.7964,  0.2757
-0.1078,  0.8437, -0.4420,  0.2177,  0.3649,  0.4028
-0.3139,  0.5595, -0.6505, -0.3161, -0.7108,  0.5546
 0.4335,  0.3986,  0.3770, -0.4932,  0.3847,  0.1810
-0.2562, -0.2894, -0.8847,  0.2633,  0.4146,  0.4036
 0.2272,  0.2966, -0.6601, -0.7011,  0.0284,  0.2778
-0.0743, -0.1421, -0.0054, -0.6770, -0.3151,  0.3597
-0.4762,  0.6891,  0.6007, -0.1467,  0.2140,  0.4266
-0.4061,  0.7193,  0.3432,  0.2669, -0.7505,  0.6147
-0.0588,  0.9731,  0.8966,  0.2902, -0.6966,  0.4955
-0.0627, -0.1439,  0.1985,  0.6999,  0.5022,  0.3077
 0.1587,  0.8494, -0.8705,  0.9827, -0.8940,  0.4263
-0.7850,  0.2473, -0.9040, -0.4308, -0.8779,  0.7199
 0.4070,  0.3369, -0.2428, -0.6236,  0.4940,  0.2215
-0.0242,  0.0513, -0.9430,  0.2885, -0.2987,  0.3947
-0.5416, -0.1322, -0.2351, -0.0604,  0.9590,  0.3683
 0.1055,  0.7783, -0.2901, -0.5090,  0.8220,  0.2984
-0.9129,  0.9015,  0.1128, -0.2473,  0.9901,  0.4776
-0.9378,  0.1424, -0.6391,  0.2619,  0.9618,  0.5368
 0.7498, -0.0963,  0.4169,  0.5549, -0.0103,  0.1614
-0.2612, -0.7156,  0.4538, -0.0460, -0.1022,  0.3717
 0.7720,  0.0552, -0.1818, -0.4622, -0.8560,  0.1685
-0.4177,  0.0070,  0.9319, -0.7812,  0.3461,  0.3052
-0.0001,  0.5542, -0.7128, -0.8336, -0.2016,  0.3803
 0.5356, -0.4194, -0.5662, -0.9666, -0.2027,  0.1776
-0.2378,  0.3187, -0.8582, -0.6948, -0.9668,  0.5474
-0.1947, -0.3579,  0.1158,  0.9869,  0.6690,  0.2992
 0.3992,  0.8365, -0.9205, -0.8593, -0.0520,  0.3154
-0.0209,  0.0793,  0.7905, -0.1067,  0.7541,  0.1864
-0.4928, -0.4524, -0.3433,  0.0951, -0.5597,  0.6261
-0.8118,  0.7404, -0.5263, -0.2280,  0.1431,  0.6349
 0.0516, -0.8480,  0.7483,  0.9023,  0.6250,  0.1959
-0.3212,  0.1093,  0.9488, -0.3766,  0.3376,  0.2735
-0.3481,  0.5490, -0.3484,  0.7797,  0.5034,  0.4379
-0.5785, -0.9170, -0.3563, -0.9258,  0.3877,  0.4121
 0.3407, -0.1391,  0.5356,  0.0720, -0.9203,  0.3458
-0.3287, -0.8954,  0.2102,  0.0241,  0.2349,  0.3247
-0.1353,  0.6954, -0.0919, -0.9692,  0.7461,  0.3338
 0.9036, -0.8982, -0.5299, -0.8733, -0.1567,  0.1187
 0.7277, -0.8368, -0.0538, -0.7489,  0.5458,  0.0830
 0.9049,  0.8878,  0.2279,  0.9470, -0.3103,  0.2194
 0.7957, -0.1308, -0.5284,  0.8817,  0.3684,  0.2172
 0.4647, -0.4931,  0.2010,  0.6292, -0.8918,  0.3371
-0.7390,  0.6849,  0.2367,  0.0626, -0.5034,  0.7039
-0.1567, -0.8711,  0.7940, -0.5932,  0.6525,  0.1710
 0.7635, -0.0265,  0.1969,  0.0545,  0.2496,  0.1445
 0.7675,  0.1354, -0.7698, -0.5460,  0.1920,  0.1728
-0.5211, -0.7372, -0.6763,  0.6897,  0.2044,  0.5217
 0.1913,  0.1980,  0.2314, -0.8816,  0.5006,  0.1998
 0.8964,  0.0694, -0.6149,  0.5059, -0.9854,  0.1825
 0.1767,  0.7104,  0.2093,  0.6452,  0.7590,  0.2832
-0.3580, -0.7541,  0.4426, -0.1193, -0.7465,  0.5657
-0.5996,  0.5766, -0.9758, -0.3933, -0.9572,  0.6800
 0.9950,  0.1641, -0.4132,  0.8579,  0.0142,  0.2003
-0.4717, -0.3894, -0.2567, -0.5111,  0.1691,  0.4266
 0.3917, -0.8561,  0.9422,  0.5061,  0.6123,  0.1212
-0.0366, -0.1087,  0.3449, -0.1025,  0.4086,  0.2475
 0.3633,  0.3943,  0.2372, -0.6980,  0.5216,  0.1925
-0.5325, -0.6466, -0.2178, -0.3589,  0.6310,  0.3568
 0.2271,  0.5200, -0.1447, -0.8011, -0.7699,  0.3128
 0.6415,  0.1993,  0.3777, -0.0178, -0.8237,  0.2181
-0.5298, -0.0768, -0.6028, -0.9490,  0.4588,  0.4356
 0.6870, -0.1431,  0.7294,  0.3141,  0.1621,  0.1632
-0.5985,  0.0591,  0.7889, -0.3900,  0.7419,  0.2945
 0.3661,  0.7984, -0.8486,  0.7572, -0.6183,  0.3449
 0.6995,  0.3342, -0.3113, -0.6972,  0.2707,  0.1712
 0.2565,  0.9126,  0.1798, -0.6043, -0.1413,  0.2893
-0.3265,  0.9839, -0.2395,  0.9854,  0.0376,  0.4770
 0.2690, -0.1722,  0.9818,  0.8599, -0.7015,  0.3954
-0.2102, -0.0768,  0.1219,  0.5607, -0.0256,  0.3949
 0.8216, -0.9555,  0.6422, -0.6231,  0.3715,  0.0801
-0.2896,  0.9484, -0.7545, -0.6249,  0.7789,  0.4370
-0.9985, -0.5448, -0.7092, -0.5931,  0.7926,  0.5402

Test data:

# synthetic_test_40.txt
#
 0.7462,  0.4006, -0.0590,  0.6543, -0.0083,  0.1935
 0.8495, -0.2260, -0.0142, -0.4911,  0.7699,  0.1078
-0.2335, -0.4049,  0.4352, -0.6183, -0.7636,  0.5088
 0.1810, -0.5142,  0.2465,  0.2767, -0.3449,  0.3136
-0.8650,  0.7611, -0.0801,  0.5277, -0.4922,  0.7140
-0.2358, -0.7466, -0.5115, -0.8413, -0.3943,  0.4533
 0.4834,  0.2300,  0.3448, -0.9832,  0.3568,  0.1360
-0.6502, -0.6300,  0.6885,  0.9652,  0.8275,  0.3046
-0.3053,  0.5604,  0.0929,  0.6329, -0.0325,  0.4756
-0.7995,  0.0740, -0.2680,  0.2086,  0.9176,  0.4565
-0.2144, -0.2141,  0.5813,  0.2902, -0.2122,  0.4119
-0.7278, -0.0987, -0.3312, -0.5641,  0.8515,  0.4438
 0.3793,  0.1976,  0.4933,  0.0839,  0.4011,  0.1905
-0.8568,  0.9573, -0.5272,  0.3212, -0.8207,  0.7415
-0.5785,  0.0056, -0.7901, -0.2223,  0.0760,  0.5551
 0.0735, -0.2188,  0.3925,  0.3570,  0.3746,  0.2191
 0.1230, -0.2838,  0.2262,  0.8715,  0.1938,  0.2878
 0.4792, -0.9248,  0.5295,  0.0366, -0.9894,  0.3149
-0.4456,  0.0697,  0.5359, -0.8938,  0.0981,  0.3879
 0.8629, -0.8505, -0.4464,  0.8385,  0.5300,  0.1769
 0.1995,  0.6659,  0.7921,  0.9454,  0.9970,  0.2330
-0.0249, -0.3066, -0.2927, -0.4923,  0.8220,  0.2437
 0.4513, -0.9481, -0.0770, -0.4374, -0.9421,  0.2879
-0.3405,  0.5931, -0.3507, -0.3842,  0.8562,  0.3987
 0.9538,  0.0471,  0.9039,  0.7760,  0.0361,  0.1706
-0.0887,  0.2104,  0.9808,  0.5478, -0.3314,  0.4128
-0.8220, -0.6302,  0.0537, -0.1658,  0.6013,  0.4306
-0.4123, -0.2880,  0.9074, -0.0461, -0.4435,  0.5144
 0.0060,  0.2867, -0.7775,  0.5161,  0.7039,  0.3599
-0.7968, -0.5484,  0.9426, -0.4308,  0.8148,  0.2979
 0.7811,  0.8450, -0.6877,  0.7594,  0.2640,  0.2362
-0.6802, -0.1113, -0.8325, -0.6694, -0.6056,  0.6544
 0.3821,  0.1476,  0.7466, -0.5107,  0.2592,  0.1648
 0.7265,  0.9683, -0.9803, -0.4943, -0.5523,  0.2454
-0.9049, -0.9797, -0.0196, -0.9090, -0.4433,  0.6447
-0.4607,  0.1811, -0.2389,  0.4050, -0.0078,  0.5229
 0.2664, -0.2932, -0.4259, -0.7336,  0.8742,  0.1834
-0.4507,  0.1029, -0.6294, -0.1158, -0.6294,  0.6081
 0.8948, -0.0124,  0.9278,  0.2899, -0.0314,  0.1534
-0.1323, -0.8813, -0.0146, -0.0697,  0.6135,  0.2386
Posted in Machine Learning | Leave a comment

Some Thoughts About C# Program Design and the “dynamic” Keyword

Posted on June 12, 2026 by jamesdmccaffrey

One night, I just couldn’t sleep. So I thought about program design and the C# “dynamic” keyword. (My life would be considered sad by most people, but I’m happy with it).

My standard machine learning regression system implemented in C# looks like:

using Sysem;

class Program
{
  static void Main(string[] args)
  {
    // all IO and control logic here
  }
  // helper methods for Main()
} 

class SomeKindRegressor
{
  // fields here
  public SomeKindRegresssor() { . . } // ctor
  public void Train() { . . }
  public double Predict() { . . }
  public double Accuracy() { . . }
  public double MSE() { . . }
  // helper methods here
}

class HelperClass
{
  // matrix inverse, etc.
}

I put the Accuracy() and MSE() methods inside the regressor class, the idea being that these two methods are essential. So, they might be called like:

. . .
model = SomeKindRegressor();
model.Train(trainX, trainY); // possibly call, display MSE()
double trainAcc = model.Accuracy(trainX, trainY, 0.10);
. . .

But, in addition to Accuracy() and MSE(), there are a lot other evaluation methods that are sometimes needed, for example RootMSE(), BaselineAccuracy(), R2(), and so on. So I decided to implement a demo that puts these secondary evaluation metrics in a separate utility class:

class Evaluation
{
  public static RootMSE(dynamic model,
    double[][] dataX, data[]Y)
  {
    . . 
    double predY = model.Predict(dataX[0]);
    . . 
    return result;
  }

  // other methods here
}

The C# “dynamic” keyword for an arbitrary object is similar to the “var” keyword for an arbitrary data type. This allows one evaluation metric, that can be used by any program-defined regression model object to be defined, instead of a different method for every regressor class. This design assumes that every regression class has a Predict() method.

Then, the evaluation methods would be called like:

model = SomeKindRegressor();
model.Train(trainX, trainY); // possibly call, display MSE()
double trainAcc = model.Accuracy(trainX, trainY, 0.10);
double trainRMSE = Evaluation.RMSE(model, trainX, trainY);
. . .

I put together a demo. The output is:

Begin C# linear regression evaluation metrics

Loading synthetic train (200) and test (40) data
Done

First three train X:
 -0.1660  0.4406 -0.9998 -0.3953 -0.7065
  0.0776 -0.1616  0.3704 -0.5911  0.7562
 -0.9452  0.3409 -0.1654  0.1174 -0.7192

First three train y:
  0.4840
  0.1568
  0.8054

Creating and training  model using SGD

Setting lrnRate = 0.0010
Setting maxEpohcs = 1000
epoch =     0  MSE =   0.1132
epoch =   200  MSE =   0.0026
epoch =   400  MSE =   0.0026
epoch =   600  MSE =   0.0026
epoch =   800  MSE =   0.0026
Done

SGD trained weights:
-0.2656  0.0333  -0.0453  0.0358  -0.1146
Bias/constant: 0.3619

Evaluating SGD model using core metrics

Accuracy train (within 0.10) = 0.4600
Accuracy test (within 0.10) = 0.6500

MSE train = 0.0026
MSE test = 0.0020

Evaluating SGD model using secondary metrics

R2 train = 0.9267
R2 test = 0.9300

Baseline accuracy train (within 0.10) = 0.1100

End demo

An interesting investigation.


I’ve always been fascinated by aircraft design. Here are three designs of U.S. fighter jets that are very beautiful to my eye. Left: North American F-86 Sabre from the 1950s. Center: Lockheed F-104 Starfighter from the 1960s. Right: General Dynamics F-16 Fighting Falcon from the 1970s (and still being made today).

Demo program. Replace “lt” (less than), “gt”, “lte”, “gte” with Boolean operator symbols (my editor chokes on symbols).

using System;
using System.IO;
using System.Collections.Generic;

namespace LinearRegressionEvaluationMetrics
{
  internal class LinearRegressionProgram
  {
    static void Main(string[] args)
    {
      Console.WriteLine("\nBegin C# linear regression" +
        " evaluation metrics ");

      // 1. load data
      Console.WriteLine("\nLoading synthetic train" +
        " (200) and test (40) data");
      string trainFile =
        "..\\..\\..\\Data\\synthetic_train_200.txt";
      int[] colsX = new int[] { 0, 1, 2, 3, 4 };
      double[][] trainX =
        MatLoad(trainFile, colsX, ',', "#");
      double[] trainY =
        MatToVec(MatLoad(trainFile,
        new int[] { 5 }, ',', "#"));

      string testFile =
        "..\\..\\..\\Data\\synthetic_test_40.txt";
      double[][] testX =
        MatLoad(testFile, colsX, ',', "#");
      double[] testY =
        MatToVec(MatLoad(testFile,
        new int[] { 5 }, ',', "#"));
      Console.WriteLine("Done ");

      Console.WriteLine("\nFirst three train X: ");
      for (int i = 0; i "lt" 3; ++i)
        VecShow(trainX[i], 4, 8);

      Console.WriteLine("\nFirst three train y: ");
      for (int i = 0; i "lt" 3; ++i)
        Console.WriteLine(trainY[i].ToString("F4").
          PadLeft(8));

      // 2. create and train model
      Console.WriteLine("\nCreating and training" +
        "  model using SGD ");

      double lrnRate = 0.001;
      int maxEpochs = 1000;
      //int seed = 0;
      Console.WriteLine("\nSetting lrnRate = " +
        lrnRate.ToString("F4"));
      Console.WriteLine("Setting maxEpohcs = " +
        maxEpochs);

      LinearRegressor model =
        new LinearRegressor();
      model.TrainSGD(trainX, trainY, lrnRate, maxEpochs);
      Console.WriteLine("Done ");

      // 2b.show model parameters
      Console.WriteLine("\nSGD trained weights: ");
      for (int i = 0; i "lt" model.weights.Length; ++i)
        Console.Write(model.weights[i].ToString("F4") + "  ");
      Console.WriteLine("\nBias/constant: " +
        model.bias.ToString("F4"));

      // 3. evaluate model
      Console.WriteLine("\nEvaluating SGD model using" +
        " core metrics ");

      double accTrain = model.Accuracy(trainX, trainY, 0.10);
      Console.WriteLine("\nAccuracy train (within 0.10) = " +
        accTrain.ToString("F4"));
      double accTest = model.Accuracy(testX, testY, 0.10);
      Console.WriteLine("Accuracy test (within 0.10) = " +
        accTest.ToString("F4"));

      double mseTrain = model.MSE(trainX, trainY);
      Console.WriteLine("\nMSE train = " +
        mseTrain.ToString("F4"));
      double mseTest = model.MSE(testX, testY);
      Console.WriteLine("MSE test = " +
        mseTest.ToString("F4"));

      Console.WriteLine("\nEvaluating SGD model using" +
        " secondary metrics ");

      double r2Train = Eval.R2(model, trainX, trainY);
      Console.WriteLine("\nR2 train = " +
        r2Train.ToString("F4"));
      double r2Test = Eval.R2(model, testX, testY);
      Console.WriteLine("R2 test = " +
        r2Test.ToString("F4"));

      double baseAccTrain =
        Eval.BaselineAccuracy(trainX, trainY, 0.10);
      Console.WriteLine("\nBaseline accuracy train " +
        "(within 0.10) = " + baseAccTrain.ToString("F4"));

      Console.WriteLine("\nEnd demo ");
      Console.ReadLine();
    } // Main

    // ------------------------------------------------------
    // helpers for Main()
    // ------------------------------------------------------

    static double[][] MatLoad(string fn, int[] usecols,
      char sep, string comment)
    {
      List"lt"double[]"gt" result = 
        new List"lt"double[]"gt"();
      string line = "";
      FileStream ifs = new FileStream(fn, FileMode.Open);
      StreamReader sr = new StreamReader(ifs);
      while ((line = sr.ReadLine()) != null)
      {
        if (line.StartsWith(comment) == true)
          continue;
        string[] tokens = line.Split(sep);
        List"lt"double"gt" lst = new List"lt"double"gt"();
        for (int j = 0; j "lt" usecols.Length; ++j)
          lst.Add(double.Parse(tokens[usecols[j]]));
        double[] row = lst.ToArray();
        result.Add(row);
      }
      sr.Close(); ifs.Close();
      return result.ToArray();
    }

    static double[] MatToVec(double[][] M)
    {
      int nRows = M.Length;
      int nCols = M[0].Length;
      double[] result = new double[nRows * nCols];
      int k = 0;
      for (int i = 0; i "lt" nRows; ++i)
        for (int j = 0; j "lt" nCols; ++j)
          result[k++] = M[i][j];
      return result;
    }

    static void VecShow(double[] vec, int dec, int wid)
    {
      for (int i = 0; i "lt" vec.Length; ++i)
        Console.Write(vec[i].ToString("F" + dec).
          PadLeft(wid));
      Console.WriteLine("");
    }
  } // class Program

  public class LinearRegressor
  {
    public double[] weights;
    public double bias;
    private Random rnd;

    public LinearRegressor(int seed = 0)
    {
      this.weights = new double[0]; // keep compiler happy
      this.bias = 0;
      this.rnd = new Random(seed);
    }

    // ------------------------------------------------------

    public void TrainSGD(double[][] trainX,
      double[] trainY, double lrnRate, int maxEpochs)
    {
      int n = trainX.Length; int dim = trainX[0].Length;
      this.weights = new double[dim];

      // initialize weights and bias
      double low = -0.01; double hi = 0.01;
      for (int i = 0; i "lt" dim; ++i)
        this.weights[i] = (hi - low) *
          this.rnd.NextDouble() + low;
      this.bias = (hi - low) *
          this.rnd.NextDouble() + low;

      int[] indices = new int[n];  // of train data
      for (int i = 0; i "lt" n; ++i)
        indices[i] = i;

      for (int epoch = 0; epoch "lt" maxEpochs; ++epoch)
      {
        this.Shuffle(indices);

        for (int i = 0; i "lt" n; ++i) // each train item
        {
          int ii = indices[i];
          double[] x = trainX[ii];
          double predY = this.Predict(x);
          double actualY = trainY[ii];
          for (int j = 0; j "lt" dim; ++j) // each weight
            this.weights[j] -= lrnRate *
              (predY - actualY) * x[j];
          this.bias -= lrnRate * (predY - actualY);
        }
        if (epoch % (int)(maxEpochs / 5) == 0) // progress
        {
          double mse = this.MSE(trainX, trainY); // core
          string s = "";
          s += "epoch = " + epoch.ToString().PadLeft(5);
          s += "  MSE = " + mse.ToString("F4").PadLeft(8);
          Console.WriteLine(s);
        }
      }
    }

    // ------------------------------------------------------

    private void Shuffle(int[] indices)
    {
      // Fisher-Yates
      for (int i = 0; i "lt" indices.Length; ++i)
      {
        int ri = this.rnd.Next(i, indices.Length);
        int tmp = indices[i];
        indices[i] = indices[ri];
        indices[ri] = tmp;
      }
    }

    // ------------------------------------------------------

    public double Predict(double[] x)
    {
      double result = 0.0;
      for (int j = 0; j "lt" x.Length; ++j)
        result += x[j] * this.weights[j];
      result += this.bias;
      return result;
    }

    // ------------------------------------------------------

    public double Accuracy(double[][] dataX, double[] dataY,
      double pctClose)
    {
      int numCorrect = 0; int numWrong = 0;
      for (int i = 0; i "lt" dataX.Length; ++i)
      {
        double actualY = dataY[i];
        double predY = this.Predict(dataX[i]);
        if (Math.Abs(predY - actualY) "lt"
          Math.Abs(pctClose * actualY))
          ++numCorrect;
        else
          ++numWrong;
      }
      return (numCorrect * 1.0) / (numWrong + numCorrect);
    }

    // ------------------------------------------------------

    public double MSE(double[][] dataX, double[] dataY)
    {
      int n = dataX.Length;
      double sum = 0.0;
      for (int i = 0; i "lt" n; ++i)
      {
        double actualY = dataY[i];
        double predY = this.Predict(dataX[i]);
        sum += (actualY - predY) * (actualY - predY);
      }
      return sum / n;
    }

  } // class LinearRegressor

  // ========================================================

  public class Eval
  {
    // secondary model evaluation metrics
    // uses 'dynamic' keyword (C# 4.0, 2010 and later)

    public static double R2(dynamic model,
      double[][] dataX, double[] dataY)
    {
      // R2 = 1.0 – (u / v)
      // where u = sum(y – y')^2 and v = sum(y – y”)^2,
      // y = actual y, y' = predicted y, y'' = mean y
      int n = dataX.Length;
      double sum = 0.0;

      for (int i = 0; i "lt" n; ++i)
        sum += dataY[i];
      double meanY = sum / n;

      double sumTop = 0.0; // u
      double sumBot = 0.0; // v
      for (int i = 0; i "lt" n; ++i)
      {
        double predY = model.Predict(dataX[i]);
        sumTop +=
          (dataY[i] - predY) * (dataY[i] - predY);
        sumBot +=
          (dataY[i] - meanY) * (dataY[i] - meanY);
      }
      return 1.0 - (sumTop / sumBot);
    }

    // ------------------------------------------------------

    public static double RootMSE(dynamic model,
      double[][] dataX, double[] dataY)
    {
      int n = dataX.Length;
      double sum = 0.0;
      for (int i = 0; i "lt" n; ++i)
      {
        double actualY = dataY[i];
        double predY = model.Predict(dataX[i]);
        sum += (actualY - predY) * (actualY - predY);
      }
      double mse = sum / n;
      return Math.Sqrt(mse);
    }

    // ------------------------------------------------------

    public static double BaselineAccuracy(double[][] dataX,
      double[] dataY, double pctClose)
    {
      // compute avg of y values
      int n = dataY.Length;
      double sum = 0.0;

      for (int i = 0; i "lt" n; ++i)
        sum += dataY[i];
      double meanY = sum / n;

      // always predict the mean
      int numCorrect = 0; int numWrong = 0;
      for (int i = 0; i "lt" dataX.Length; ++i)
      {
        double actualY = dataY[i];
        double predY = meanY;
        if (Math.Abs(predY - actualY) "lt"
          Math.Abs(pctClose * actualY))
          ++numCorrect;
        else
          ++numWrong;
      }
      return (numCorrect * 1.0) / (numWrong + numCorrect);
    }

    // ------------------------------------------------------

    public static double BaselineMSE(double[][] dataX,
      double[] dataY)
    {
      // compute avg of y values
      int n = dataY.Length;
      double sum = 0.0;

      for (int i = 0; i "lt" n; ++i)
        sum += dataY[i];
      double meanY = sum / n;

      // always predict the mean
      sum = 0.0;
      for (int i = 0; i "lt" n; ++i)
      {
        double actualY = dataY[i];
        double predY = meanY;
        sum += (actualY - predY) * (actualY - predY);
      }
      return sum / n;
    }

    // ------------------------------------------------------

  } // class Eval

} // ns

Trainining data:

# synthetic_train_200.txt
#
-0.1660,  0.4406, -0.9998, -0.3953, -0.7065,  0.4840
 0.0776, -0.1616,  0.3704, -0.5911,  0.7562,  0.1568
-0.9452,  0.3409, -0.1654,  0.1174, -0.7192,  0.8054
 0.9365, -0.3732,  0.3846,  0.7528,  0.7892,  0.1345
-0.8299, -0.9219, -0.6603,  0.7563, -0.8033,  0.7955
 0.0663,  0.3838, -0.3690,  0.3730,  0.6693,  0.3206
-0.9634,  0.5003,  0.9777,  0.4963, -0.4391,  0.7377
-0.1042,  0.8172, -0.4128, -0.4244, -0.7399,  0.4801
-0.9613,  0.3577, -0.5767, -0.4689, -0.0169,  0.6861
-0.7065,  0.1786,  0.3995, -0.7953, -0.1719,  0.5569
 0.3888, -0.1716, -0.9001,  0.0718,  0.3276,  0.2500
 0.1731,  0.8068, -0.7251, -0.7214,  0.6148,  0.3297
-0.2046, -0.6693,  0.8550, -0.3045,  0.5016,  0.2129
 0.2473,  0.5019, -0.3022, -0.4601,  0.7918,  0.2613
-0.1438,  0.9297,  0.3269,  0.2434, -0.7705,  0.5171
 0.1568, -0.1837, -0.5259,  0.8068,  0.1474,  0.3307
-0.9943,  0.2343, -0.3467,  0.0541,  0.7719,  0.5581
 0.2467, -0.9684,  0.8589,  0.3818,  0.9946,  0.1092
-0.6553, -0.7257,  0.8652,  0.3936, -0.8680,  0.7018
 0.8460,  0.4230, -0.7515, -0.9602, -0.9476,  0.1996
-0.9434, -0.5076,  0.7201,  0.0777,  0.1056,  0.5664
 0.9392,  0.1221, -0.9627,  0.6013, -0.5341,  0.1533
 0.6142, -0.2243,  0.7271,  0.4942,  0.1125,  0.1661
 0.4260,  0.1194, -0.9749, -0.8561,  0.9346,  0.2230
 0.1362, -0.5934, -0.4953,  0.4877, -0.6091,  0.3810
 0.6937, -0.5203, -0.0125,  0.2399,  0.6580,  0.1460
-0.6864, -0.9628, -0.8600, -0.0273,  0.2127,  0.5387
 0.9772,  0.1595, -0.2397,  0.1019,  0.4907,  0.1611
 0.3385, -0.4702, -0.8673, -0.2598,  0.2594,  0.2270
-0.8669, -0.4794,  0.6095, -0.6131,  0.2789,  0.4700
 0.0493,  0.8496, -0.4734, -0.8681,  0.4701,  0.3516
 0.8639, -0.9721, -0.5313,  0.2336,  0.8980,  0.1412
 0.9004,  0.1133,  0.8312,  0.2831, -0.2200,  0.1782
 0.0991,  0.8524,  0.8375, -0.2102,  0.9265,  0.2150
-0.6521, -0.7473, -0.7298,  0.0113, -0.9570,  0.7422
 0.6190, -0.3105,  0.8802,  0.1640,  0.7577,  0.1056
 0.6895,  0.8108, -0.0802,  0.0927,  0.5972,  0.2214
 0.1982, -0.9689,  0.1870, -0.1326,  0.6147,  0.1310
-0.3695,  0.7858,  0.1557, -0.6320,  0.5759,  0.3773
-0.1596,  0.3581,  0.8372, -0.9992,  0.9535,  0.2071
-0.2468,  0.9476,  0.2094,  0.6577,  0.1494,  0.4132
 0.1737,  0.5000,  0.7166,  0.5102,  0.3961,  0.2611
 0.7290, -0.3546,  0.3416, -0.0983, -0.2358,  0.1332
-0.3652,  0.2438, -0.1395,  0.9476,  0.3556,  0.4170
-0.6029, -0.1466, -0.3133,  0.5953,  0.7600,  0.4334
-0.4596, -0.4953,  0.7098,  0.0554,  0.6043,  0.2775
 0.1450,  0.4663,  0.0380,  0.5418,  0.1377,  0.2931
-0.8636, -0.2442, -0.8407,  0.9656, -0.6368,  0.7429
 0.6237,  0.7499,  0.3768,  0.1390, -0.6781,  0.2185
-0.5499,  0.1850, -0.3755,  0.8326,  0.8193,  0.4399
-0.4858, -0.7782, -0.6141, -0.0008,  0.4572,  0.4197
 0.7033, -0.1683,  0.2334, -0.5327, -0.7961,  0.1776
 0.0317, -0.0457, -0.6947,  0.2436,  0.0880,  0.3345
 0.5031, -0.5559,  0.0387,  0.5706, -0.9553,  0.3107
-0.3513,  0.7458,  0.6894,  0.0769,  0.7332,  0.3170
 0.2205,  0.5992, -0.9309,  0.5405,  0.4635,  0.3532
-0.4806, -0.4859,  0.2646, -0.3094,  0.5932,  0.3202
 0.9809, -0.3995, -0.7140,  0.8026,  0.0831,  0.1600
 0.9495,  0.2732,  0.9878,  0.0921,  0.0529,  0.1289
-0.9476, -0.6792,  0.4913, -0.9392, -0.2669,  0.5966
 0.7247,  0.3854,  0.3819, -0.6227, -0.1162,  0.1550
-0.5922, -0.5045, -0.4757,  0.5003, -0.0860,  0.5863
-0.8861,  0.0170, -0.5761,  0.5972, -0.4053,  0.7301
 0.6877, -0.2380,  0.4997,  0.0223,  0.0819,  0.1404
 0.9189,  0.6079, -0.9354,  0.4188, -0.0700,  0.1907
-0.1428, -0.7820,  0.2676,  0.6059,  0.3936,  0.2790
 0.5324, -0.3151,  0.6917, -0.1425,  0.6480,  0.1071
-0.8432, -0.9633, -0.8666, -0.0828, -0.7733,  0.7784
-0.9444,  0.5097, -0.2103,  0.4939, -0.0952,  0.6787
-0.0520,  0.6063, -0.1952,  0.8094, -0.9259,  0.4836
 0.5477, -0.7487,  0.2370, -0.9793,  0.0773,  0.1241
 0.2450,  0.8116,  0.9799,  0.4222,  0.4636,  0.2355
 0.8186, -0.1983, -0.5003, -0.6531, -0.7611,  0.1511
-0.4714,  0.6382, -0.3788,  0.9648, -0.4667,  0.5950
 0.0673, -0.3711,  0.8215, -0.2669, -0.1328,  0.2677
-0.9381,  0.4338,  0.7820, -0.9454,  0.0441,  0.5518
-0.3480,  0.7190,  0.1170,  0.3805, -0.0943,  0.4724
-0.9813,  0.1535, -0.3771,  0.0345,  0.8328,  0.5438
-0.1471, -0.5052, -0.2574,  0.8637,  0.8737,  0.3042
-0.5454, -0.3712, -0.6505,  0.2142, -0.1728,  0.5783
 0.6327, -0.6297,  0.4038, -0.5193,  0.1484,  0.1153
-0.5424,  0.3282, -0.0055,  0.0380, -0.6506,  0.6613
 0.1414,  0.9935,  0.6337,  0.1887,  0.9520,  0.2540
-0.9351, -0.8128, -0.8693, -0.0965, -0.2491,  0.7353
 0.9507, -0.6640,  0.9456,  0.5349,  0.6485,  0.1059
-0.0462, -0.9737, -0.2940, -0.0159,  0.4602,  0.2606
-0.0627, -0.0852, -0.7247, -0.9782,  0.5166,  0.2977
 0.0478,  0.5098, -0.0723, -0.7504, -0.3750,  0.3335
 0.0090,  0.3477,  0.5403, -0.7393, -0.9542,  0.4415
-0.9748,  0.3449,  0.3736, -0.1015,  0.8296,  0.4358
 0.2887, -0.9895, -0.0311,  0.7186,  0.6608,  0.2057
 0.1570, -0.4518,  0.1211,  0.3435, -0.2951,  0.3244
 0.7117, -0.6099,  0.4946, -0.4208,  0.5476,  0.1096
-0.2929, -0.5726,  0.5346, -0.3827,  0.4665,  0.2465
 0.4889, -0.5572, -0.5718, -0.6021, -0.7150,  0.2163
-0.7782,  0.3491,  0.5996, -0.8389, -0.5366,  0.6516
-0.5847,  0.8347,  0.4226,  0.1078, -0.3910,  0.6134
 0.8469,  0.4121, -0.0439, -0.7476,  0.9521,  0.1571
-0.6803, -0.5948, -0.1376, -0.1916, -0.7065,  0.7156
 0.2878,  0.5086, -0.5785,  0.2019,  0.4979,  0.2980
 0.2764,  0.1943, -0.4090,  0.4632,  0.8906,  0.2960
-0.8877,  0.6705, -0.6155, -0.2098, -0.3998,  0.7107
-0.8398,  0.8093, -0.2597,  0.0614, -0.0118,  0.6502
-0.8476,  0.0158, -0.4769, -0.2859, -0.7839,  0.7715
 0.5751, -0.7868,  0.9714, -0.6457,  0.1448,  0.1175
 0.4802, -0.7001,  0.1022, -0.5668,  0.5184,  0.1090
 0.4458, -0.6469,  0.7239, -0.9604,  0.7205,  0.0779
 0.5175,  0.4339,  0.9747, -0.4438, -0.9924,  0.2879
 0.8678,  0.7158,  0.4577,  0.0334,  0.4139,  0.1678
 0.5406,  0.5012,  0.2264, -0.1963,  0.3946,  0.2088
-0.9938,  0.5498,  0.7928, -0.5214, -0.7585,  0.7687
 0.7661,  0.0863, -0.4266, -0.7233, -0.4197,  0.1466
 0.2277, -0.3517, -0.0853, -0.1118,  0.6563,  0.1767
 0.3499, -0.5570, -0.0655, -0.3705,  0.2537,  0.1632
 0.7547, -0.1046,  0.5689, -0.0861,  0.3125,  0.1257
 0.8186,  0.2110,  0.5335,  0.0094, -0.0039,  0.1391
 0.6858, -0.8644,  0.1465,  0.8855,  0.0357,  0.1845
-0.4967,  0.4015,  0.0805,  0.8977,  0.2487,  0.4663
 0.6760, -0.9841,  0.9787, -0.8446, -0.3557,  0.1509
-0.1203, -0.4885,  0.6054, -0.0443, -0.7313,  0.4854
 0.8557,  0.7919, -0.0169,  0.7134, -0.1628,  0.2002
 0.0115, -0.6209,  0.9300, -0.4116, -0.7931,  0.4052
-0.7114, -0.9718,  0.4319,  0.1290,  0.5892,  0.3661
 0.3915,  0.5557, -0.1870,  0.2955, -0.6404,  0.2954
-0.3564, -0.6548, -0.1827, -0.5172, -0.1862,  0.4622
 0.2392, -0.4959,  0.5857, -0.1341, -0.2850,  0.2470
-0.3394,  0.3947, -0.4627,  0.6166, -0.4094,  0.5325
 0.7107,  0.7768, -0.6312,  0.1707,  0.7964,  0.2757
-0.1078,  0.8437, -0.4420,  0.2177,  0.3649,  0.4028
-0.3139,  0.5595, -0.6505, -0.3161, -0.7108,  0.5546
 0.4335,  0.3986,  0.3770, -0.4932,  0.3847,  0.1810
-0.2562, -0.2894, -0.8847,  0.2633,  0.4146,  0.4036
 0.2272,  0.2966, -0.6601, -0.7011,  0.0284,  0.2778
-0.0743, -0.1421, -0.0054, -0.6770, -0.3151,  0.3597
-0.4762,  0.6891,  0.6007, -0.1467,  0.2140,  0.4266
-0.4061,  0.7193,  0.3432,  0.2669, -0.7505,  0.6147
-0.0588,  0.9731,  0.8966,  0.2902, -0.6966,  0.4955
-0.0627, -0.1439,  0.1985,  0.6999,  0.5022,  0.3077
 0.1587,  0.8494, -0.8705,  0.9827, -0.8940,  0.4263
-0.7850,  0.2473, -0.9040, -0.4308, -0.8779,  0.7199
 0.4070,  0.3369, -0.2428, -0.6236,  0.4940,  0.2215
-0.0242,  0.0513, -0.9430,  0.2885, -0.2987,  0.3947
-0.5416, -0.1322, -0.2351, -0.0604,  0.9590,  0.3683
 0.1055,  0.7783, -0.2901, -0.5090,  0.8220,  0.2984
-0.9129,  0.9015,  0.1128, -0.2473,  0.9901,  0.4776
-0.9378,  0.1424, -0.6391,  0.2619,  0.9618,  0.5368
 0.7498, -0.0963,  0.4169,  0.5549, -0.0103,  0.1614
-0.2612, -0.7156,  0.4538, -0.0460, -0.1022,  0.3717
 0.7720,  0.0552, -0.1818, -0.4622, -0.8560,  0.1685
-0.4177,  0.0070,  0.9319, -0.7812,  0.3461,  0.3052
-0.0001,  0.5542, -0.7128, -0.8336, -0.2016,  0.3803
 0.5356, -0.4194, -0.5662, -0.9666, -0.2027,  0.1776
-0.2378,  0.3187, -0.8582, -0.6948, -0.9668,  0.5474
-0.1947, -0.3579,  0.1158,  0.9869,  0.6690,  0.2992
 0.3992,  0.8365, -0.9205, -0.8593, -0.0520,  0.3154
-0.0209,  0.0793,  0.7905, -0.1067,  0.7541,  0.1864
-0.4928, -0.4524, -0.3433,  0.0951, -0.5597,  0.6261
-0.8118,  0.7404, -0.5263, -0.2280,  0.1431,  0.6349
 0.0516, -0.8480,  0.7483,  0.9023,  0.6250,  0.1959
-0.3212,  0.1093,  0.9488, -0.3766,  0.3376,  0.2735
-0.3481,  0.5490, -0.3484,  0.7797,  0.5034,  0.4379
-0.5785, -0.9170, -0.3563, -0.9258,  0.3877,  0.4121
 0.3407, -0.1391,  0.5356,  0.0720, -0.9203,  0.3458
-0.3287, -0.8954,  0.2102,  0.0241,  0.2349,  0.3247
-0.1353,  0.6954, -0.0919, -0.9692,  0.7461,  0.3338
 0.9036, -0.8982, -0.5299, -0.8733, -0.1567,  0.1187
 0.7277, -0.8368, -0.0538, -0.7489,  0.5458,  0.0830
 0.9049,  0.8878,  0.2279,  0.9470, -0.3103,  0.2194
 0.7957, -0.1308, -0.5284,  0.8817,  0.3684,  0.2172
 0.4647, -0.4931,  0.2010,  0.6292, -0.8918,  0.3371
-0.7390,  0.6849,  0.2367,  0.0626, -0.5034,  0.7039
-0.1567, -0.8711,  0.7940, -0.5932,  0.6525,  0.1710
 0.7635, -0.0265,  0.1969,  0.0545,  0.2496,  0.1445
 0.7675,  0.1354, -0.7698, -0.5460,  0.1920,  0.1728
-0.5211, -0.7372, -0.6763,  0.6897,  0.2044,  0.5217
 0.1913,  0.1980,  0.2314, -0.8816,  0.5006,  0.1998
 0.8964,  0.0694, -0.6149,  0.5059, -0.9854,  0.1825
 0.1767,  0.7104,  0.2093,  0.6452,  0.7590,  0.2832
-0.3580, -0.7541,  0.4426, -0.1193, -0.7465,  0.5657
-0.5996,  0.5766, -0.9758, -0.3933, -0.9572,  0.6800
 0.9950,  0.1641, -0.4132,  0.8579,  0.0142,  0.2003
-0.4717, -0.3894, -0.2567, -0.5111,  0.1691,  0.4266
 0.3917, -0.8561,  0.9422,  0.5061,  0.6123,  0.1212
-0.0366, -0.1087,  0.3449, -0.1025,  0.4086,  0.2475
 0.3633,  0.3943,  0.2372, -0.6980,  0.5216,  0.1925
-0.5325, -0.6466, -0.2178, -0.3589,  0.6310,  0.3568
 0.2271,  0.5200, -0.1447, -0.8011, -0.7699,  0.3128
 0.6415,  0.1993,  0.3777, -0.0178, -0.8237,  0.2181
-0.5298, -0.0768, -0.6028, -0.9490,  0.4588,  0.4356
 0.6870, -0.1431,  0.7294,  0.3141,  0.1621,  0.1632
-0.5985,  0.0591,  0.7889, -0.3900,  0.7419,  0.2945
 0.3661,  0.7984, -0.8486,  0.7572, -0.6183,  0.3449
 0.6995,  0.3342, -0.3113, -0.6972,  0.2707,  0.1712
 0.2565,  0.9126,  0.1798, -0.6043, -0.1413,  0.2893
-0.3265,  0.9839, -0.2395,  0.9854,  0.0376,  0.4770
 0.2690, -0.1722,  0.9818,  0.8599, -0.7015,  0.3954
-0.2102, -0.0768,  0.1219,  0.5607, -0.0256,  0.3949
 0.8216, -0.9555,  0.6422, -0.6231,  0.3715,  0.0801
-0.2896,  0.9484, -0.7545, -0.6249,  0.7789,  0.4370
-0.9985, -0.5448, -0.7092, -0.5931,  0.7926,  0.5402

Test data:

# synthetic_test_40.txt
#
 0.7462,  0.4006, -0.0590,  0.6543, -0.0083,  0.1935
 0.8495, -0.2260, -0.0142, -0.4911,  0.7699,  0.1078
-0.2335, -0.4049,  0.4352, -0.6183, -0.7636,  0.5088
 0.1810, -0.5142,  0.2465,  0.2767, -0.3449,  0.3136
-0.8650,  0.7611, -0.0801,  0.5277, -0.4922,  0.7140
-0.2358, -0.7466, -0.5115, -0.8413, -0.3943,  0.4533
 0.4834,  0.2300,  0.3448, -0.9832,  0.3568,  0.1360
-0.6502, -0.6300,  0.6885,  0.9652,  0.8275,  0.3046
-0.3053,  0.5604,  0.0929,  0.6329, -0.0325,  0.4756
-0.7995,  0.0740, -0.2680,  0.2086,  0.9176,  0.4565
-0.2144, -0.2141,  0.5813,  0.2902, -0.2122,  0.4119
-0.7278, -0.0987, -0.3312, -0.5641,  0.8515,  0.4438
 0.3793,  0.1976,  0.4933,  0.0839,  0.4011,  0.1905
-0.8568,  0.9573, -0.5272,  0.3212, -0.8207,  0.7415
-0.5785,  0.0056, -0.7901, -0.2223,  0.0760,  0.5551
 0.0735, -0.2188,  0.3925,  0.3570,  0.3746,  0.2191
 0.1230, -0.2838,  0.2262,  0.8715,  0.1938,  0.2878
 0.4792, -0.9248,  0.5295,  0.0366, -0.9894,  0.3149
-0.4456,  0.0697,  0.5359, -0.8938,  0.0981,  0.3879
 0.8629, -0.8505, -0.4464,  0.8385,  0.5300,  0.1769
 0.1995,  0.6659,  0.7921,  0.9454,  0.9970,  0.2330
-0.0249, -0.3066, -0.2927, -0.4923,  0.8220,  0.2437
 0.4513, -0.9481, -0.0770, -0.4374, -0.9421,  0.2879
-0.3405,  0.5931, -0.3507, -0.3842,  0.8562,  0.3987
 0.9538,  0.0471,  0.9039,  0.7760,  0.0361,  0.1706
-0.0887,  0.2104,  0.9808,  0.5478, -0.3314,  0.4128
-0.8220, -0.6302,  0.0537, -0.1658,  0.6013,  0.4306
-0.4123, -0.2880,  0.9074, -0.0461, -0.4435,  0.5144
 0.0060,  0.2867, -0.7775,  0.5161,  0.7039,  0.3599
-0.7968, -0.5484,  0.9426, -0.4308,  0.8148,  0.2979
 0.7811,  0.8450, -0.6877,  0.7594,  0.2640,  0.2362
-0.6802, -0.1113, -0.8325, -0.6694, -0.6056,  0.6544
 0.3821,  0.1476,  0.7466, -0.5107,  0.2592,  0.1648
 0.7265,  0.9683, -0.9803, -0.4943, -0.5523,  0.2454
-0.9049, -0.9797, -0.0196, -0.9090, -0.4433,  0.6447
-0.4607,  0.1811, -0.2389,  0.4050, -0.0078,  0.5229
 0.2664, -0.2932, -0.4259, -0.7336,  0.8742,  0.1834
-0.4507,  0.1029, -0.6294, -0.1158, -0.6294,  0.6081
 0.8948, -0.0124,  0.9278,  0.2899, -0.0314,  0.1534
-0.1323, -0.8813, -0.0146, -0.0697,  0.6135,  0.2386
Posted in Machine Learning | Leave a comment

“AdaBoost.R2 Regression Using C#” in Visual Studio Magazine

Posted on June 11, 2026 by jamesdmccaffrey

I wrote an article titled “AdaBoost.R2 Regression Using C#” in the June 2026 edition of Microsoft Visual Studio Magazine. See https://visualstudiomagazine.com/articles/2026/06/01/adaboost-r2-regression-using-csharp.aspx.

The goal of a machine learning regression problem is to predict a single numeric value. For example, a bank might want to predict the maximum safe loan amount for a customer based on age, account balance, current debt, and so on. Regression techniques fall into two categories: those based on decision tree structures, and those that use non-tree techniques.

In a nutshell:

* AdaBoost.R2 regression is a machine learning technique used to predict a single numeric value
* AdaBoost.R2 builds a sequence of decision tree regressors where each accepted tree improves prediction compared to earlier trees.
* Final AdaBoost.R2 predictions are produced as a weighted median across the decision tree predictions.

Examples of non-tree techniques include linear regression, quadratic regression, kernel ridge regression, nearest neighbors regression, and neural network regression. The three most common tree-based regression techniques are random forest regression (which includes a variant called bagging tree regression), AdaBoost.R2 regression, and gradient boost regression. Each regression technique has pros and cons. If there was one regression technique that worked best for all types of datasets, there would only be one technique, not many.

My article presents a complete demo of AdaBoost.R2 implemented using the C# language. AdaBoost.R2 regression sequentially creates an ensemble (collection) of simple decision trees, where each tree is a bit better at prediction than the previous tree. For a given input x, the predicted y value is the weighted median of the predictions of the collection of decision trees.

The output of the article demo program is:

Begin AdaBoost.R2 (tree) regression from scratch C#

Loading synthetic train (200) and test (40) data

First three train X:
 -0.1660  0.4406 -0.9998 -0.3953 -0.7065
  0.0776 -0.1616  0.3704 -0.5911  0.7562
 -0.9452  0.3409 -0.1654  0.1174 -0.7192

First three train y:
  0.4840
  0.1568
  0.8054

Setting maxLearners = 100
Setting tree maxDepth = 5
Setting tree minSamples = 2
Setting tree minLeaf = 1

Training AdaBoost.R2 model
Done
Created 100 learners

Accuracy train (within 0.10): 0.8250
Accuracy test (within 0.10): 0.5250

MSE train: 0.0004
MSE test: 0.0022

Predicting for x =
 -0.1660  0.4406 -0.9998 -0.3953 -0.7065
Predicted y = 0.4890

End demo

I based my demo program directly on the 1997 source research paper, “Improving Regressors Using Boosting Techniques” (1997), by H. Drucker.

A weight value is assigned to each training data item, where the weight values sum to 1. Initially all training item weights are the same. For the demo data, because there are 200 items, each initial data item weight is 1/200 = 0.005.

Each training data item is fed to the current decision tree and an average loss/error for the current tree is computed. The average loss for the current tree is used to compute a beta value for the tree. A small value for beta means high confidence in the current tree, and vice versa.

The weights associated with each training data item are updated using the beta value for the current tree and the loss/error for each data item. At this point, a new set of training data items is probabilistically created from the source data items, where data items with high error have a greater probability of being selected than data items with small error.

The new training data items are fed to the next decision tree. Because the new training dataset has high-error items, the new tree will concentrate on learning to predict those difficult-to-predict items.

Each new training dataset is constructed probabilistically, and so there is a chance that a newly created decision tree is not better than the previous decision tree. In this case, the newly created decision tree is not added to the collection of trees. This is why the actual number of trees created may be less than the maxLearners parameter.

AdaBoost.R2 regression is not used nearly as often as random forest regression, or gradient boost regression. My AdaBoost.R2 demo is therefore most likely to be used with legacy systems that already use AdaBoost.R2 regression.


Like most of my friends, I get obsessed with all kinds of strange things. I have been obsessed with machine learning for many years.

For reasons that are unknown to me, I am moderately obsessed with movies that feature a creature that makes noises that are translated to “chittering” in the closed caption text. The are an amazing number of chittering instances in movies, but you’ll usually only notice them if you’re specifically looking for them.

Left: In the fantasy movie “The Dark Crystal” (192), Gelfling Jen and his girlfriend Kira, must overthrow the evil, ruling Skeksis by restoring a powerful broken Crystal. Early in the movie, Jen travels throw a creepy swamp with a chittering creature.

Right: In the fantasy movie “The Legend of Ochi” (2025), farm girl Yuri discovers a wounded baby primate-like creature and works to return it to its family. She is menaced by a hunting party led by her father and her stepbrother. When Yuri goes through a cave system, some insects are chittering while a bat watches.

Posted in Machine Learning | Leave a comment

An Example of RANSAC Regression Using Scikit on the Diabetes Dataset

Posted on June 10, 2026 by jamesdmccaffrey

The RANSAC (random sample consensus) regression technique can be used when the training dataset is known to have outliers (typically, blatant errors). RANSAC is a meta-heuristic that can be applied to any base regression technique, but is usually applied to basic linear regression.

The idea is:

inliers = []
loop many times
  take a very small sample from train data
  train a model
  use model on train data to find inliers
  if enough inliers
    train a "good" model on inliers
  end-if
end-loop
return best model found

I usually think of RANSAC as a kind of automatic outlier detection and removal, combined with regression. Before I go any further, let me note that the original RANSAC idea was developed in 1980 for a very specific image processing problem scenario, and the idea works quite well. But RANSAC has been applied to more general regression scenarios, and I am not a fan of RANSAC in those general scenarios.

I put together a RANSAC regression demo using the scikit-learn RANSACRegressor module on the Diabetes Dataset. The output of the demo:

Begin scikit RANSAC (random sample consensus)
 on Diabetes Dataset demo

Loading diabetes train (342), test (100) data
Done

First three X predictors:
[0.5900 1.0000 0.3210 0.1010 0.1570 0.0932 0.3800 0.4000
 0.4860 0.0870]
[0.4800 0.0000 0.2160 0.0870 0.1830 0.1032 0.7000 0.3000
 0.3892 0.0690]
[0.7200 1.0000 0.3050 0.0930 0.1560 0.0936 0.4100 0.4000
 0.4673 0.0850]

First three y targets:
0.1510
0.0750
0.1410

Creating scikit RANSAC model
Done

Training RANSAC model
Done.

Evaluating model

Accuracy (within 0.10) train = 0.1462
Accuracy (within 0.10) test = 0.1413

MSE train = 0.0040
MSE test = 0.0049

===============================

Creating, training base linear regression model

LR accuracy (within 0.10) train = 0.1813
LR accuracy (within 0.10) test = 0.2500

LR MSE train = 0.0029
LR MSE test = 0.0028

End demo

The RANSAC results were poor, as expected, and the RANSAC results were even worse than just plain linear regression, as expected.

The raw Diabetes Dataset looks like:

59, 2, 32.1, 101.00, 157,  93.2, 38, 4.00, 4.8598, 87, 151
48, 1, 21.6,  87.00, 183, 103.2, 70, 3.00, 3.8918, 69,  75
72, 2, 30.5,  93.00, 156,  93.6, 41, 4.00, 4.6728, 85, 141
. . .

Each line represents a patient. The first 10 values on each line are predictors. The last value on each line is the target value (a diabetes metric) to predict. The predictors are: age, sex, body mass index, blood pressure, serum cholesterol, low-density lipoproteins, high-density lipoproteins, total cholesterol, triglycerides, blood sugar. There are 442 data items.

The sex encoding isn’t explained anywhere but I suspect male = 1, female = 2 because there are 235 1 values and 206 2 values).

Note that this Diabetes Dataset, which is included as an example dataset in the Python language scikit-learn library, is not the same as the Pima Diabetes Dataset from the UCI dataset repository. See https://jamesmccaffreyblog.com/2026/02/03/the-origin-and-history-of-scikit-learn-diabetes-dataset/.


I tracked down, and read, the original 1980 RANSAC paper.

I converted the sex values from 1,2 into 0,1. Then I applied divide-by-constant normalization by dividing the 10 predictor columns by (100, 1, 100, 1000, 1000, 1000, 100, 10, 10, 1000) and the target y values by 1000. The resulting encoded and normalized data looks like:

0.5900, 1.0000, 0.3210, . . . 0.1510
0.4800, 0.0000, 0.2160, . . . 0.0750
0.7200, 1.0000, 0.3050, . . . 0.1410
. . .

I split the 442-items into a 342-item training set and a 100-item test set. I have done many experiments with the Diabetes Dataset and I’ve concluded the the default target value in the last column (a patient diabetes score) simply cannot be predicted well. But the variables in columns [4], [5], [6], [7], and [8] can be meaningfully predicted from the other columns.


In some sense, machine learning regression can be thought of as searching for hidden patterns in data.

Every cover of Playboy Magazine, except for the very first one (December 1953), has the company bunny logo somewhere. On most covers, the logo is clear and easy to see. But on some covers, the logo cleverly hidden. Looking for hidden items in a photo has an odd appeal to most people.

Left: On the cover of the issue from February 1993, the bunny logo is very difficult to spot. It’s incorporated into the long red threads on the model’s clothing. I added a white circle for you. Right: One of the most difficult hidden logos ever: on the cover of the July 1993 issue, the bunny logo is disguised as part of the water reflections. I circled it in red for you.

Demo program. Replace “lt” (less than) in the accuracy() function with the Boolean less-than operator symbol. (My blog editor chokes on symbols).

# diabetes_scikit_ransac.py
# RANSAC (random sample consensus)
# for the Diabetes Dataset

import numpy as np
from sklearn.linear_model import RANSACRegressor

# RANSACRegressor(estimator=None, *, min_samples=None,
# residual_threshold=None, is_data_valid=None,
# is_model_valid=None, max_trials=100, max_skips=inf,
# stop_n_inliers=inf, stop_score=inf,
# stop_probability=0.99, loss='absolute_error',
# random_state=None)

# -----------------------------------------------------------

np.set_printoptions(precision=4, suppress=True,
  floatmode='fixed', linewidth=60)

# -----------------------------------------------------------

def accuracy(model, data_X, data_y, pct_close):
  n = len(data_X)
  n_correct = 0; n_wrong = 0
  for i in range(n):
    x = data_X[i].reshape(1,-1)
    y = data_y[i]
    y_pred = model.predict(x)[0]

    if np.abs(y - y_pred) "lt" np.abs(y * pct_close):
      n_correct += 1
    else: 
      n_wrong += 1
  return n_correct / (n_correct + n_wrong)

def mse(model, data_X, data_y):
  n = len(data_X)
  sum = 0.0
  for i in range(n):
    actual_y = data_y[i]
    pred_y = model.predict(data_X[i].reshape(1, -1))[0]
    diff = actual_y - pred_y
    sum += diff * diff
  return sum /n

# -----------------------------------------------------------

print("\nBegin scikit RANSAC (random sample consensus) " +\
 "on Diabetes Dataset demo ")

print("\nLoading diabetes train (342), test (100) data ")
train_file = ".\\Data\\diabetes_norm_train_342.txt"

cols_X = [0,1,2,3,4,5,6,7,8,9]  
col_y = 10  # cols # 4 5 6 7 8 are much better
train_X = np.loadtxt(train_file, comments="#",
  usecols=cols_X, delimiter=",",  dtype=np.float64)
train_y = np.loadtxt(train_file, comments="#",
  usecols=col_y, delimiter=",",  dtype=np.float64)

test_file = ".\\Data\\diabetes_norm_test_100.txt"
test_X = np.loadtxt(test_file, comments="#",
  usecols=cols_X, delimiter=",",  dtype=np.float64)
test_y = np.loadtxt(test_file, comments="#",
  usecols=col_y, delimiter=",",  dtype=np.float64)
print("Done ")

# alternative normalization and split
# from sklearn.datasets import load_diabetes
# from sklearn.model_selection import train_test_split
# X, y = load_diabetes(return_X_y=True, scaled=True)
# train_X, test_X, train_y, test_y = \
#   train_test_split(X, y, random_state=0)  # 25% test

print("\nFirst three X predictors: ")
for i in range(3):
  print(train_X[i])
print("\nFirst three y targets: ")
for i in range(3):
  print("%0.4f" % train_y[i])

print("\nCreating scikit RANSAC model ")
model = RANSACRegressor(random_state=0)
print("Done ")

print("\nTraining RANSAC model ")
model.fit(train_X, train_y)
print("Done. ")

print("\nEvaluating model ")
acc_train = accuracy(model, train_X, train_y, 0.10)
acc_test = accuracy(model, test_X, test_y, 0.10)
print("\nAccuracy (within 0.10) train = %0.4f " % \
  acc_train)
print("Accuracy (within 0.10) test = %0.4f " % \
  acc_test)

mse_train = mse(model, train_X, train_y)
mse_test = mse(model, test_X, test_y)
print("\nMSE train = %0.4f " % mse_train)
print("MSE test = %0.4f " % mse_test)

print("\n=============================== ")

print("\nCreating, training base linear regression model")
from sklearn.linear_model import LinearRegression
lr_model = LinearRegression()
lr_model.fit(train_X, train_y)

acc_train = accuracy(lr_model, train_X, train_y, 0.10)
acc_test = accuracy(lr_model, test_X, test_y, 0.10)
print("\nLR accuracy (within 0.10) train = %0.4f " % \
  acc_train)
print("LR accuracy (within 0.10) test = %0.4f " % \
  acc_test)

mse_train = mse(lr_model, train_X, train_y)
mse_test = mse(lr_model, test_X, test_y)
print("\nLR MSE train = %0.4f " % mse_train)
print("LR MSE test = %0.4f " % mse_test)

print("\nEnd demo ")

Training data:


# diabetes_norm_train_342.txt
# cols [0] to [9] predictors. col [10] target
# norm division constants:
# 100, -1, 100, 1000, 1000, 1000, 100, 10, 10, 1000, 1000
#
0.5900, 1.0000, 0.3210, 0.1010, 0.1570, 0.0932, 0.3800, 0.4000, 0.4860, 0.0870, 0.1510
0.4800, 0.0000, 0.2160, 0.0870, 0.1830, 0.1032, 0.7000, 0.3000, 0.3892, 0.0690, 0.0750
0.7200, 1.0000, 0.3050, 0.0930, 0.1560, 0.0936, 0.4100, 0.4000, 0.4673, 0.0850, 0.1410
0.2400, 0.0000, 0.2530, 0.0840, 0.1980, 0.1314, 0.4000, 0.5000, 0.4890, 0.0890, 0.2060
0.5000, 0.0000, 0.2300, 0.1010, 0.1920, 0.1254, 0.5200, 0.4000, 0.4291, 0.0800, 0.1350
0.2300, 0.0000, 0.2260, 0.0890, 0.1390, 0.0648, 0.6100, 0.2000, 0.4190, 0.0680, 0.0970
0.3600, 1.0000, 0.2200, 0.0900, 0.1600, 0.0996, 0.5000, 0.3000, 0.3951, 0.0820, 0.1380
0.6600, 1.0000, 0.2620, 0.1140, 0.2550, 0.1850, 0.5600, 0.4550, 0.4249, 0.0920, 0.0630
0.6000, 1.0000, 0.3210, 0.0830, 0.1790, 0.1194, 0.4200, 0.4000, 0.4477, 0.0940, 0.1100
0.2900, 0.0000, 0.3000, 0.0850, 0.1800, 0.0934, 0.4300, 0.4000, 0.5385, 0.0880, 0.3100
0.2200, 0.0000, 0.1860, 0.0970, 0.1140, 0.0576, 0.4600, 0.2000, 0.3951, 0.0830, 0.1010
0.5600, 1.0000, 0.2800, 0.0850, 0.1840, 0.1448, 0.3200, 0.6000, 0.3584, 0.0770, 0.0690
0.5300, 0.0000, 0.2370, 0.0920, 0.1860, 0.1092, 0.6200, 0.3000, 0.4304, 0.0810, 0.1790
0.5000, 1.0000, 0.2620, 0.0970, 0.1860, 0.1054, 0.4900, 0.4000, 0.5063, 0.0880, 0.1850
0.6100, 0.0000, 0.2400, 0.0910, 0.2020, 0.1154, 0.7200, 0.3000, 0.4291, 0.0730, 0.1180
0.3400, 1.0000, 0.2470, 0.1180, 0.2540, 0.1842, 0.3900, 0.7000, 0.5037, 0.0810, 0.1710
0.4700, 0.0000, 0.3030, 0.1090, 0.2070, 0.1002, 0.7000, 0.3000, 0.5215, 0.0980, 0.1660
0.6800, 1.0000, 0.2750, 0.1110, 0.2140, 0.1470, 0.3900, 0.5000, 0.4942, 0.0910, 0.1440
0.3800, 0.0000, 0.2540, 0.0840, 0.1620, 0.1030, 0.4200, 0.4000, 0.4443, 0.0870, 0.0970
0.4100, 0.0000, 0.2470, 0.0830, 0.1870, 0.1082, 0.6000, 0.3000, 0.4543, 0.0780, 0.1680
0.3500, 0.0000, 0.2110, 0.0820, 0.1560, 0.0878, 0.5000, 0.3000, 0.4511, 0.0950, 0.0680
0.2500, 1.0000, 0.2430, 0.0950, 0.1620, 0.0986, 0.5400, 0.3000, 0.3850, 0.0870, 0.0490
0.2500, 0.0000, 0.2600, 0.0920, 0.1870, 0.1204, 0.5600, 0.3000, 0.3970, 0.0880, 0.0680
0.6100, 1.0000, 0.3200, 0.1037, 0.2100, 0.0852, 0.3500, 0.6000, 0.6107, 0.1240, 0.2450
0.3100, 0.0000, 0.2970, 0.0880, 0.1670, 0.1034, 0.4800, 0.4000, 0.4357, 0.0780, 0.1840
0.3000, 1.0000, 0.2520, 0.0830, 0.1780, 0.1184, 0.3400, 0.5000, 0.4852, 0.0830, 0.2020
0.1900, 0.0000, 0.1920, 0.0870, 0.1240, 0.0540, 0.5700, 0.2000, 0.4174, 0.0900, 0.1370
0.4200, 0.0000, 0.3190, 0.0830, 0.1580, 0.0876, 0.5300, 0.3000, 0.4466, 0.1010, 0.0850
0.6300, 0.0000, 0.2440, 0.0730, 0.1600, 0.0914, 0.4800, 0.3000, 0.4635, 0.0780, 0.1310
0.6700, 1.0000, 0.2580, 0.1130, 0.1580, 0.0542, 0.6400, 0.2000, 0.5293, 0.1040, 0.2830
0.3200, 0.0000, 0.3050, 0.0890, 0.1820, 0.1106, 0.5600, 0.3000, 0.4344, 0.0890, 0.1290
0.4200, 0.0000, 0.2030, 0.0710, 0.1610, 0.0812, 0.6600, 0.2000, 0.4234, 0.0810, 0.0590
0.5800, 1.0000, 0.3800, 0.1030, 0.1500, 0.1072, 0.2200, 0.7000, 0.4644, 0.0980, 0.3410
0.5700, 0.0000, 0.2170, 0.0940, 0.1570, 0.0580, 0.8200, 0.2000, 0.4443, 0.0920, 0.0870
0.5300, 0.0000, 0.2050, 0.0780, 0.1470, 0.0842, 0.5200, 0.3000, 0.3989, 0.0750, 0.0650
0.6200, 1.0000, 0.2350, 0.0803, 0.2250, 0.1128, 0.8600, 0.2620, 0.4875, 0.0960, 0.1020
0.5200, 0.0000, 0.2850, 0.1100, 0.1950, 0.0972, 0.6000, 0.3000, 0.5242, 0.0850, 0.2650
0.4600, 0.0000, 0.2740, 0.0780, 0.1710, 0.0880, 0.5800, 0.3000, 0.4828, 0.0900, 0.2760
0.4800, 1.0000, 0.3300, 0.1230, 0.2530, 0.1636, 0.4400, 0.6000, 0.5425, 0.0970, 0.2520
0.4800, 1.0000, 0.2770, 0.0730, 0.1910, 0.1194, 0.4600, 0.4000, 0.4852, 0.0920, 0.0900
0.5000, 1.0000, 0.2560, 0.1010, 0.2290, 0.1622, 0.4300, 0.5000, 0.4779, 0.1140, 0.1000
0.2100, 0.0000, 0.2010, 0.0630, 0.1350, 0.0690, 0.5400, 0.3000, 0.4094, 0.0890, 0.0550
0.3200, 1.0000, 0.2540, 0.0903, 0.1530, 0.1004, 0.3400, 0.4500, 0.4533, 0.0830, 0.0610
0.5400, 0.0000, 0.2420, 0.0740, 0.2040, 0.1090, 0.8200, 0.2000, 0.4174, 0.1090, 0.0920
0.6100, 1.0000, 0.3270, 0.0970, 0.1770, 0.1184, 0.2900, 0.6000, 0.4997, 0.0870, 0.2590
0.5600, 1.0000, 0.2310, 0.1040, 0.1810, 0.1164, 0.4700, 0.4000, 0.4477, 0.0790, 0.0530
0.3300, 0.0000, 0.2530, 0.0850, 0.1550, 0.0850, 0.5100, 0.3000, 0.4554, 0.0700, 0.1900
0.2700, 0.0000, 0.1960, 0.0780, 0.1280, 0.0680, 0.4300, 0.3000, 0.4443, 0.0710, 0.1420
0.6700, 1.0000, 0.2250, 0.0980, 0.1910, 0.1192, 0.6100, 0.3000, 0.3989, 0.0860, 0.0750
0.3700, 1.0000, 0.2770, 0.0930, 0.1800, 0.1194, 0.3000, 0.6000, 0.5030, 0.0880, 0.1420
0.5800, 0.0000, 0.2570, 0.0990, 0.1570, 0.0916, 0.4900, 0.3000, 0.4407, 0.0930, 0.1550
0.6500, 1.0000, 0.2790, 0.1030, 0.1590, 0.0968, 0.4200, 0.4000, 0.4615, 0.0860, 0.2250
0.3400, 0.0000, 0.2550, 0.0930, 0.2180, 0.1440, 0.5700, 0.4000, 0.4443, 0.0880, 0.0590
0.4600, 0.0000, 0.2490, 0.1150, 0.1980, 0.1296, 0.5400, 0.4000, 0.4277, 0.1030, 0.1040
0.3500, 0.0000, 0.2870, 0.0970, 0.2040, 0.1268, 0.6400, 0.3000, 0.4190, 0.0930, 0.1820
0.3700, 0.0000, 0.2180, 0.0840, 0.1840, 0.1010, 0.7300, 0.3000, 0.3912, 0.0930, 0.1280
0.3700, 0.0000, 0.3020, 0.0870, 0.1660, 0.0960, 0.4000, 0.4150, 0.5011, 0.0870, 0.0520
0.4100, 0.0000, 0.2050, 0.0800, 0.1240, 0.0488, 0.6400, 0.2000, 0.4025, 0.0750, 0.0370
0.6000, 0.0000, 0.2040, 0.1050, 0.1980, 0.0784, 0.9900, 0.2000, 0.4635, 0.0790, 0.1700
0.6600, 1.0000, 0.2400, 0.0980, 0.2360, 0.1464, 0.5800, 0.4000, 0.5063, 0.0960, 0.1700
0.2900, 0.0000, 0.2600, 0.0830, 0.1410, 0.0652, 0.6400, 0.2000, 0.4078, 0.0830, 0.0610
0.3700, 1.0000, 0.2680, 0.0790, 0.1570, 0.0980, 0.2800, 0.6000, 0.5043, 0.0960, 0.1440
0.4100, 1.0000, 0.2570, 0.0830, 0.1810, 0.1066, 0.6600, 0.3000, 0.3738, 0.0850, 0.0520
0.3900, 0.0000, 0.2290, 0.0770, 0.2040, 0.1432, 0.4600, 0.4000, 0.4304, 0.0740, 0.1280
0.6700, 1.0000, 0.2400, 0.0830, 0.1430, 0.0772, 0.4900, 0.3000, 0.4431, 0.0940, 0.0710
0.3600, 1.0000, 0.2410, 0.1120, 0.1930, 0.1250, 0.3500, 0.6000, 0.5106, 0.0950, 0.1630
0.4600, 1.0000, 0.2470, 0.0850, 0.1740, 0.1232, 0.3000, 0.6000, 0.4644, 0.0960, 0.1500
0.6000, 1.0000, 0.2500, 0.0897, 0.1850, 0.1208, 0.4600, 0.4020, 0.4511, 0.0920, 0.0970
0.5900, 1.0000, 0.2360, 0.0830, 0.1650, 0.1000, 0.4700, 0.4000, 0.4500, 0.0920, 0.1600
0.5300, 0.0000, 0.2210, 0.0930, 0.1340, 0.0762, 0.4600, 0.3000, 0.4078, 0.0960, 0.1780
0.4800, 0.0000, 0.1990, 0.0910, 0.1890, 0.1096, 0.6900, 0.3000, 0.3951, 0.1010, 0.0480
0.4800, 0.0000, 0.2950, 0.1310, 0.2070, 0.1322, 0.4700, 0.4000, 0.4935, 0.1060, 0.2700
0.6600, 1.0000, 0.2600, 0.0910, 0.2640, 0.1466, 0.6500, 0.4000, 0.5568, 0.0870, 0.2020
0.5200, 1.0000, 0.2450, 0.0940, 0.2170, 0.1494, 0.4800, 0.5000, 0.4585, 0.0890, 0.1110
0.5200, 1.0000, 0.2660, 0.1110, 0.2090, 0.1264, 0.6100, 0.3000, 0.4682, 0.1090, 0.0850
0.4600, 1.0000, 0.2350, 0.0870, 0.1810, 0.1148, 0.4400, 0.4000, 0.4710, 0.0980, 0.0420
0.4000, 1.0000, 0.2900, 0.1150, 0.0970, 0.0472, 0.3500, 0.2770, 0.4304, 0.0950, 0.1700
0.2200, 0.0000, 0.2300, 0.0730, 0.1610, 0.0978, 0.5400, 0.3000, 0.3829, 0.0910, 0.2000
0.5000, 0.0000, 0.2100, 0.0880, 0.1400, 0.0718, 0.3500, 0.4000, 0.5112, 0.0710, 0.2520
0.2000, 0.0000, 0.2290, 0.0870, 0.1910, 0.1282, 0.5300, 0.4000, 0.3892, 0.0850, 0.1130
0.6800, 0.0000, 0.2750, 0.1070, 0.2410, 0.1496, 0.6400, 0.4000, 0.4920, 0.0900, 0.1430
0.5200, 1.0000, 0.2430, 0.0860, 0.1970, 0.1336, 0.4400, 0.5000, 0.4575, 0.0910, 0.0510
0.4400, 0.0000, 0.2310, 0.0870, 0.2130, 0.1264, 0.7700, 0.3000, 0.3871, 0.0720, 0.0520
0.3800, 0.0000, 0.2730, 0.0810, 0.1460, 0.0816, 0.4700, 0.3000, 0.4466, 0.0810, 0.2100
0.4900, 0.0000, 0.2270, 0.0653, 0.1680, 0.0962, 0.6200, 0.2710, 0.3892, 0.0600, 0.0650
0.6100, 0.0000, 0.3300, 0.0950, 0.1820, 0.1148, 0.5400, 0.3000, 0.4190, 0.0740, 0.1410
0.2900, 1.0000, 0.1940, 0.0830, 0.1520, 0.1058, 0.3900, 0.4000, 0.3584, 0.0830, 0.0550
0.6100, 0.0000, 0.2580, 0.0980, 0.2350, 0.1258, 0.7600, 0.3000, 0.5112, 0.0820, 0.1340
0.3400, 1.0000, 0.2260, 0.0750, 0.1660, 0.0918, 0.6000, 0.3000, 0.4263, 0.1080, 0.0420
0.3600, 0.0000, 0.2190, 0.0890, 0.1890, 0.1052, 0.6800, 0.3000, 0.4369, 0.0960, 0.1110
0.5200, 0.0000, 0.2400, 0.0830, 0.1670, 0.0866, 0.7100, 0.2000, 0.3850, 0.0940, 0.0980
0.6100, 0.0000, 0.3120, 0.0790, 0.2350, 0.1568, 0.4700, 0.5000, 0.5050, 0.0960, 0.1640
0.4300, 0.0000, 0.2680, 0.1230, 0.1930, 0.1022, 0.6700, 0.3000, 0.4779, 0.0940, 0.0480
0.3500, 0.0000, 0.2040, 0.0650, 0.1870, 0.1056, 0.6700, 0.2790, 0.4277, 0.0780, 0.0960
0.2700, 0.0000, 0.2480, 0.0910, 0.1890, 0.1068, 0.6900, 0.3000, 0.4190, 0.0690, 0.0900
0.2900, 0.0000, 0.2100, 0.0710, 0.1560, 0.0970, 0.3800, 0.4000, 0.4654, 0.0900, 0.1620
0.6400, 1.0000, 0.2730, 0.1090, 0.1860, 0.1076, 0.3800, 0.5000, 0.5308, 0.0990, 0.1500
0.4100, 0.0000, 0.3460, 0.0873, 0.2050, 0.1426, 0.4100, 0.5000, 0.4673, 0.1100, 0.2790
0.4900, 1.0000, 0.2590, 0.0910, 0.1780, 0.1066, 0.5200, 0.3000, 0.4575, 0.0750, 0.0920
0.4800, 0.0000, 0.2040, 0.0980, 0.2090, 0.1394, 0.4600, 0.5000, 0.4771, 0.0780, 0.0830
0.5300, 0.0000, 0.2800, 0.0880, 0.2330, 0.1438, 0.5800, 0.4000, 0.5050, 0.0910, 0.1280
0.5300, 1.0000, 0.2220, 0.1130, 0.1970, 0.1152, 0.6700, 0.3000, 0.4304, 0.1000, 0.1020
0.2300, 0.0000, 0.2900, 0.0900, 0.2160, 0.1314, 0.6500, 0.3000, 0.4585, 0.0910, 0.3020
0.6500, 1.0000, 0.3020, 0.0980, 0.2190, 0.1606, 0.4000, 0.5000, 0.4522, 0.0840, 0.1980
0.4100, 0.0000, 0.3240, 0.0940, 0.1710, 0.1044, 0.5600, 0.3000, 0.3970, 0.0760, 0.0950
0.5500, 1.0000, 0.2340, 0.0830, 0.1660, 0.1016, 0.4600, 0.4000, 0.4522, 0.0960, 0.0530
0.2200, 0.0000, 0.1930, 0.0820, 0.1560, 0.0932, 0.5200, 0.3000, 0.3989, 0.0710, 0.1340
0.5600, 0.0000, 0.3100, 0.0787, 0.1870, 0.1414, 0.3400, 0.5500, 0.4060, 0.0900, 0.1440
0.5400, 1.0000, 0.3060, 0.1033, 0.1440, 0.0798, 0.3000, 0.4800, 0.5142, 0.1010, 0.2320
0.5900, 1.0000, 0.2550, 0.0953, 0.1900, 0.1394, 0.3500, 0.5430, 0.4357, 0.1170, 0.0810
0.6000, 1.0000, 0.2340, 0.0880, 0.1530, 0.0898, 0.5800, 0.3000, 0.3258, 0.0950, 0.1040
0.5400, 0.0000, 0.2680, 0.0870, 0.2060, 0.1220, 0.6800, 0.3000, 0.4382, 0.0800, 0.0590
0.2500, 0.0000, 0.2830, 0.0870, 0.1930, 0.1280, 0.4900, 0.4000, 0.4382, 0.0920, 0.2460
0.5400, 1.0000, 0.2770, 0.1130, 0.2000, 0.1284, 0.3700, 0.5000, 0.5153, 0.1130, 0.2970
0.5500, 0.0000, 0.3660, 0.1130, 0.1990, 0.0944, 0.4300, 0.4630, 0.5730, 0.0970, 0.2580
0.4000, 1.0000, 0.2650, 0.0930, 0.2360, 0.1470, 0.3700, 0.7000, 0.5561, 0.0920, 0.2290
0.6200, 1.0000, 0.3180, 0.1150, 0.1990, 0.1286, 0.4400, 0.5000, 0.4883, 0.0980, 0.2750
0.6500, 0.0000, 0.2440, 0.1200, 0.2220, 0.1356, 0.3700, 0.6000, 0.5509, 0.1240, 0.2810
0.3300, 1.0000, 0.2540, 0.1020, 0.2060, 0.1410, 0.3900, 0.5000, 0.4868, 0.1050, 0.1790
0.5300, 0.0000, 0.2200, 0.0940, 0.1750, 0.0880, 0.5900, 0.3000, 0.4942, 0.0980, 0.2000
0.3500, 0.0000, 0.2680, 0.0980, 0.1620, 0.1036, 0.4500, 0.4000, 0.4205, 0.0860, 0.2000
0.6600, 0.0000, 0.2800, 0.1010, 0.1950, 0.1292, 0.4000, 0.5000, 0.4860, 0.0940, 0.1730
0.6200, 1.0000, 0.3390, 0.1010, 0.2210, 0.1564, 0.3500, 0.6000, 0.4997, 0.1030, 0.1800
0.5000, 1.0000, 0.2960, 0.0943, 0.3000, 0.2424, 0.3300, 0.9090, 0.4812, 0.1090, 0.0840
0.4700, 0.0000, 0.2860, 0.0970, 0.1640, 0.0906, 0.5600, 0.3000, 0.4466, 0.0880, 0.1210
0.4700, 1.0000, 0.2560, 0.0940, 0.1650, 0.0748, 0.4000, 0.4000, 0.5526, 0.0930, 0.1610
0.2400, 0.0000, 0.2070, 0.0870, 0.1490, 0.0806, 0.6100, 0.2000, 0.3611, 0.0780, 0.0990
0.5800, 1.0000, 0.2620, 0.0910, 0.2170, 0.1242, 0.7100, 0.3000, 0.4691, 0.0680, 0.1090
0.3400, 0.0000, 0.2060, 0.0870, 0.1850, 0.1122, 0.5800, 0.3000, 0.4304, 0.0740, 0.1150
0.5100, 0.0000, 0.2790, 0.0960, 0.1960, 0.1222, 0.4200, 0.5000, 0.5069, 0.1200, 0.2680
0.3100, 1.0000, 0.3530, 0.1250, 0.1870, 0.1124, 0.4800, 0.4000, 0.4890, 0.1090, 0.2740
0.2200, 0.0000, 0.1990, 0.0750, 0.1750, 0.1086, 0.5400, 0.3000, 0.4127, 0.0720, 0.1580
0.5300, 1.0000, 0.2440, 0.0920, 0.2140, 0.1460, 0.5000, 0.4000, 0.4500, 0.0970, 0.1070
0.3700, 1.0000, 0.2140, 0.0830, 0.1280, 0.0696, 0.4900, 0.3000, 0.3850, 0.0840, 0.0830
0.2800, 0.0000, 0.3040, 0.0850, 0.1980, 0.1156, 0.6700, 0.3000, 0.4344, 0.0800, 0.1030
0.4700, 0.0000, 0.3160, 0.0840, 0.1540, 0.0880, 0.3000, 0.5100, 0.5199, 0.1050, 0.2720
0.2300, 0.0000, 0.1880, 0.0780, 0.1450, 0.0720, 0.6300, 0.2000, 0.3912, 0.0860, 0.0850
0.5000, 0.0000, 0.3100, 0.1230, 0.1780, 0.1050, 0.4800, 0.4000, 0.4828, 0.0880, 0.2800
0.5800, 1.0000, 0.3670, 0.1170, 0.1660, 0.0938, 0.4400, 0.4000, 0.4949, 0.1090, 0.3360
0.5500, 0.0000, 0.3210, 0.1100, 0.1640, 0.0842, 0.4200, 0.4000, 0.5242, 0.0900, 0.2810
0.6000, 1.0000, 0.2770, 0.1070, 0.1670, 0.1146, 0.3800, 0.4000, 0.4277, 0.0950, 0.1180
0.4100, 0.0000, 0.3080, 0.0810, 0.2140, 0.1520, 0.2800, 0.7600, 0.5136, 0.1230, 0.3170
0.6000, 1.0000, 0.2750, 0.1060, 0.2290, 0.1438, 0.5100, 0.4000, 0.5142, 0.0910, 0.2350
0.4000, 0.0000, 0.2690, 0.0920, 0.2030, 0.1198, 0.7000, 0.3000, 0.4190, 0.0810, 0.0600
0.5700, 1.0000, 0.3070, 0.0900, 0.2040, 0.1478, 0.3400, 0.6000, 0.4710, 0.0930, 0.1740
0.3700, 0.0000, 0.3830, 0.1130, 0.1650, 0.0946, 0.5300, 0.3000, 0.4466, 0.0790, 0.2590
0.4000, 1.0000, 0.3190, 0.0950, 0.1980, 0.1356, 0.3800, 0.5000, 0.4804, 0.0930, 0.1780
0.3300, 0.0000, 0.3500, 0.0890, 0.2000, 0.1304, 0.4200, 0.4760, 0.4927, 0.1010, 0.1280
0.3200, 1.0000, 0.2780, 0.0890, 0.2160, 0.1462, 0.5500, 0.4000, 0.4304, 0.0910, 0.0960
0.3500, 1.0000, 0.2590, 0.0810, 0.1740, 0.1024, 0.3100, 0.6000, 0.5313, 0.0820, 0.1260
0.5500, 0.0000, 0.3290, 0.1020, 0.1640, 0.1062, 0.4100, 0.4000, 0.4431, 0.0890, 0.2880
0.4900, 0.0000, 0.2600, 0.0930, 0.1830, 0.1002, 0.6400, 0.3000, 0.4543, 0.0880, 0.0880
0.3900, 1.0000, 0.2630, 0.1150, 0.2180, 0.1582, 0.3200, 0.7000, 0.4935, 0.1090, 0.2920
0.6000, 1.0000, 0.2230, 0.1130, 0.1860, 0.1258, 0.4600, 0.4000, 0.4263, 0.0940, 0.0710
0.6700, 1.0000, 0.2830, 0.0930, 0.2040, 0.1322, 0.4900, 0.4000, 0.4736, 0.0920, 0.1970
0.4100, 1.0000, 0.3200, 0.1090, 0.2510, 0.1706, 0.4900, 0.5000, 0.5056, 0.1030, 0.1860
0.4400, 0.0000, 0.2540, 0.0950, 0.1620, 0.0926, 0.5300, 0.3000, 0.4407, 0.0830, 0.0250
0.4800, 1.0000, 0.2330, 0.0893, 0.2120, 0.1428, 0.4600, 0.4610, 0.4754, 0.0980, 0.0840
0.4500, 0.0000, 0.2030, 0.0743, 0.1900, 0.1262, 0.4900, 0.3880, 0.4304, 0.0790, 0.0960
0.4700, 0.0000, 0.3040, 0.1200, 0.1990, 0.1200, 0.4600, 0.4000, 0.5106, 0.0870, 0.1950
0.4600, 0.0000, 0.2060, 0.0730, 0.1720, 0.1070, 0.5100, 0.3000, 0.4249, 0.0800, 0.0530
0.3600, 1.0000, 0.3230, 0.1150, 0.2860, 0.1994, 0.3900, 0.7000, 0.5472, 0.1120, 0.2170
0.3400, 0.0000, 0.2920, 0.0730, 0.1720, 0.1082, 0.4900, 0.4000, 0.4304, 0.0910, 0.1720
0.5300, 1.0000, 0.3310, 0.1170, 0.1830, 0.1190, 0.4800, 0.4000, 0.4382, 0.1060, 0.1310
0.6100, 0.0000, 0.2460, 0.1010, 0.2090, 0.1068, 0.7700, 0.3000, 0.4836, 0.0880, 0.2140
0.3700, 0.0000, 0.2020, 0.0810, 0.1620, 0.0878, 0.6300, 0.3000, 0.4025, 0.0880, 0.0590
0.3300, 1.0000, 0.2080, 0.0840, 0.1250, 0.0702, 0.4600, 0.3000, 0.3784, 0.0660, 0.0700
0.6800, 0.0000, 0.3280, 0.1057, 0.2050, 0.1164, 0.4000, 0.5130, 0.5493, 0.1170, 0.2200
0.4900, 1.0000, 0.3190, 0.0940, 0.2340, 0.1558, 0.3400, 0.7000, 0.5398, 0.1220, 0.2680
0.4800, 0.0000, 0.2390, 0.1090, 0.2320, 0.1052, 0.3700, 0.6000, 0.6107, 0.0960, 0.1520
0.5500, 1.0000, 0.2450, 0.0840, 0.1790, 0.1058, 0.6600, 0.3000, 0.3584, 0.0870, 0.0470
0.4300, 0.0000, 0.2210, 0.0660, 0.1340, 0.0772, 0.4500, 0.3000, 0.4078, 0.0800, 0.0740
0.6000, 1.0000, 0.3300, 0.0970, 0.2170, 0.1256, 0.4500, 0.5000, 0.5447, 0.1120, 0.2950
0.3100, 1.0000, 0.1900, 0.0930, 0.1370, 0.0730, 0.4700, 0.3000, 0.4443, 0.0780, 0.1010
0.5300, 1.0000, 0.2730, 0.0820, 0.1190, 0.0550, 0.3900, 0.3000, 0.4828, 0.0930, 0.1510
0.6700, 0.0000, 0.2280, 0.0870, 0.1660, 0.0986, 0.5200, 0.3000, 0.4344, 0.0920, 0.1270
0.6100, 1.0000, 0.2820, 0.1060, 0.2040, 0.1320, 0.5200, 0.4000, 0.4605, 0.0960, 0.2370
0.6200, 0.0000, 0.2890, 0.0873, 0.2060, 0.1272, 0.3300, 0.6240, 0.5434, 0.0990, 0.2250
0.6000, 0.0000, 0.2560, 0.0870, 0.2070, 0.1258, 0.6900, 0.3000, 0.4111, 0.0840, 0.0810
0.4200, 0.0000, 0.2490, 0.0910, 0.2040, 0.1418, 0.3800, 0.5000, 0.4796, 0.0890, 0.1510
0.3800, 1.0000, 0.2680, 0.1050, 0.1810, 0.1192, 0.3700, 0.5000, 0.4820, 0.0910, 0.1070
0.6200, 0.0000, 0.2240, 0.0790, 0.2220, 0.1474, 0.5900, 0.4000, 0.4357, 0.0760, 0.0640
0.6100, 1.0000, 0.2690, 0.1110, 0.2360, 0.1724, 0.3900, 0.6000, 0.4812, 0.0890, 0.1380
0.6100, 1.0000, 0.2310, 0.1130, 0.1860, 0.1144, 0.4700, 0.4000, 0.4812, 0.1050, 0.1850
0.5300, 0.0000, 0.2860, 0.0880, 0.1710, 0.0988, 0.4100, 0.4000, 0.5050, 0.0990, 0.2650
0.2800, 1.0000, 0.2470, 0.0970, 0.1750, 0.0996, 0.3200, 0.5000, 0.5380, 0.0870, 0.1010
0.2600, 1.0000, 0.3030, 0.0890, 0.2180, 0.1522, 0.3100, 0.7000, 0.5159, 0.0820, 0.1370
0.3000, 0.0000, 0.2130, 0.0870, 0.1340, 0.0630, 0.6300, 0.2000, 0.3689, 0.0660, 0.1430
0.5000, 0.0000, 0.2610, 0.1090, 0.2430, 0.1606, 0.6200, 0.4000, 0.4625, 0.0890, 0.1410
0.4800, 0.0000, 0.2020, 0.0950, 0.1870, 0.1174, 0.5300, 0.4000, 0.4419, 0.0850, 0.0790
0.5100, 0.0000, 0.2520, 0.1030, 0.1760, 0.1122, 0.3700, 0.5000, 0.4898, 0.0900, 0.2920
0.4700, 1.0000, 0.2250, 0.0820, 0.1310, 0.0668, 0.4100, 0.3000, 0.4754, 0.0890, 0.1780
0.6400, 1.0000, 0.2350, 0.0970, 0.2030, 0.1290, 0.5900, 0.3000, 0.4318, 0.0770, 0.0910
0.5100, 1.0000, 0.2590, 0.0760, 0.2400, 0.1690, 0.3900, 0.6000, 0.5075, 0.0960, 0.1160
0.3000, 0.0000, 0.2090, 0.1040, 0.1520, 0.0838, 0.4700, 0.3000, 0.4663, 0.0970, 0.0860
0.5600, 1.0000, 0.2870, 0.0990, 0.2080, 0.1464, 0.3900, 0.5000, 0.4727, 0.0970, 0.1220
0.4200, 0.0000, 0.2210, 0.0850, 0.2130, 0.1386, 0.6000, 0.4000, 0.4277, 0.0940, 0.0720
0.6200, 1.0000, 0.2670, 0.1150, 0.1830, 0.1240, 0.3500, 0.5000, 0.4788, 0.1000, 0.1290
0.3400, 0.0000, 0.3140, 0.0870, 0.1490, 0.0938, 0.4600, 0.3000, 0.3829, 0.0770, 0.1420
0.6000, 0.0000, 0.2220, 0.1047, 0.2210, 0.1054, 0.6000, 0.3680, 0.5628, 0.0930, 0.0900
0.6400, 0.0000, 0.2100, 0.0923, 0.2270, 0.1468, 0.6500, 0.3490, 0.4331, 0.1020, 0.1580
0.3900, 1.0000, 0.2120, 0.0900, 0.1820, 0.1104, 0.6000, 0.3000, 0.4060, 0.0980, 0.0390
0.7100, 1.0000, 0.2650, 0.1050, 0.2810, 0.1736, 0.5500, 0.5000, 0.5568, 0.0840, 0.1960
0.4800, 1.0000, 0.2920, 0.1100, 0.2180, 0.1516, 0.3900, 0.6000, 0.4920, 0.0980, 0.2220
0.7900, 1.0000, 0.2700, 0.1030, 0.1690, 0.1108, 0.3700, 0.5000, 0.4663, 0.1100, 0.2770
0.4000, 0.0000, 0.3070, 0.0990, 0.1770, 0.0854, 0.5000, 0.4000, 0.5338, 0.0850, 0.0990
0.4900, 1.0000, 0.2880, 0.0920, 0.2070, 0.1400, 0.4400, 0.5000, 0.4745, 0.0920, 0.1960
0.5100, 0.0000, 0.3060, 0.1030, 0.1980, 0.1066, 0.5700, 0.3000, 0.5148, 0.1000, 0.2020
0.5700, 0.0000, 0.3010, 0.1170, 0.2020, 0.1396, 0.4200, 0.5000, 0.4625, 0.1200, 0.1550
0.5900, 1.0000, 0.2470, 0.1140, 0.1520, 0.1048, 0.2900, 0.5000, 0.4511, 0.0880, 0.0770
0.5100, 0.0000, 0.2770, 0.0990, 0.2290, 0.1456, 0.6900, 0.3000, 0.4277, 0.0770, 0.1910
0.7400, 0.0000, 0.2980, 0.1010, 0.1710, 0.1048, 0.5000, 0.3000, 0.4394, 0.0860, 0.0700
0.6700, 0.0000, 0.2670, 0.1050, 0.2250, 0.1354, 0.6900, 0.3000, 0.4635, 0.0960, 0.0730
0.4900, 0.0000, 0.1980, 0.0880, 0.1880, 0.1148, 0.5700, 0.3000, 0.4394, 0.0930, 0.0490
0.5700, 0.0000, 0.2330, 0.0880, 0.1550, 0.0636, 0.7800, 0.2000, 0.4205, 0.0780, 0.0650
0.5600, 1.0000, 0.3510, 0.1230, 0.1640, 0.0950, 0.3800, 0.4000, 0.5043, 0.1170, 0.2630
0.5200, 1.0000, 0.2970, 0.1090, 0.2280, 0.1628, 0.3100, 0.8000, 0.5142, 0.1030, 0.2480
0.6900, 0.0000, 0.2930, 0.1240, 0.2230, 0.1390, 0.5400, 0.4000, 0.5011, 0.1020, 0.2960
0.3700, 0.0000, 0.2030, 0.0830, 0.1850, 0.1246, 0.3800, 0.5000, 0.4719, 0.0880, 0.2140
0.2400, 0.0000, 0.2250, 0.0890, 0.1410, 0.0680, 0.5200, 0.3000, 0.4654, 0.0840, 0.1850
0.5500, 1.0000, 0.2270, 0.0930, 0.1540, 0.0942, 0.5300, 0.3000, 0.3526, 0.0750, 0.0780
0.3600, 0.0000, 0.2280, 0.0870, 0.1780, 0.1160, 0.4100, 0.4000, 0.4654, 0.0820, 0.0930
0.4200, 1.0000, 0.2400, 0.1070, 0.1500, 0.0850, 0.4400, 0.3000, 0.4654, 0.0960, 0.2520
0.2100, 0.0000, 0.2420, 0.0760, 0.1470, 0.0770, 0.5300, 0.3000, 0.4443, 0.0790, 0.1500
0.4100, 0.0000, 0.2020, 0.0620, 0.1530, 0.0890, 0.5000, 0.3000, 0.4249, 0.0890, 0.0770
0.5700, 1.0000, 0.2940, 0.1090, 0.1600, 0.0876, 0.3100, 0.5000, 0.5333, 0.0920, 0.2080
0.2000, 1.0000, 0.2210, 0.0870, 0.1710, 0.0996, 0.5800, 0.3000, 0.4205, 0.0780, 0.0770
0.6700, 1.0000, 0.2360, 0.1113, 0.1890, 0.1054, 0.7000, 0.2700, 0.4220, 0.0930, 0.1080
0.3400, 0.0000, 0.2520, 0.0770, 0.1890, 0.1206, 0.5300, 0.4000, 0.4344, 0.0790, 0.1600
0.4100, 1.0000, 0.2490, 0.0860, 0.1920, 0.1150, 0.6100, 0.3000, 0.4382, 0.0940, 0.0530
0.3800, 1.0000, 0.3300, 0.0780, 0.3010, 0.2150, 0.5000, 0.6020, 0.5193, 0.1080, 0.2200
0.5100, 0.0000, 0.2350, 0.1010, 0.1950, 0.1210, 0.5100, 0.4000, 0.4745, 0.0940, 0.1540
0.5200, 1.0000, 0.2640, 0.0913, 0.2180, 0.1520, 0.3900, 0.5590, 0.4905, 0.0990, 0.2590
0.6700, 0.0000, 0.2980, 0.0800, 0.1720, 0.0934, 0.6300, 0.3000, 0.4357, 0.0820, 0.0900
0.6100, 0.0000, 0.3000, 0.1080, 0.1940, 0.1000, 0.5200, 0.3730, 0.5347, 0.1050, 0.2460
0.6700, 1.0000, 0.2500, 0.1117, 0.1460, 0.0934, 0.3300, 0.4420, 0.4585, 0.1030, 0.1240
0.5600, 0.0000, 0.2700, 0.1050, 0.2470, 0.1606, 0.5400, 0.5000, 0.5088, 0.0940, 0.0670
0.6400, 0.0000, 0.2000, 0.0747, 0.1890, 0.1148, 0.6200, 0.3050, 0.4111, 0.0910, 0.0720
0.5800, 1.0000, 0.2550, 0.1120, 0.1630, 0.1106, 0.2900, 0.6000, 0.4762, 0.0860, 0.2570
0.5500, 0.0000, 0.2820, 0.0910, 0.2500, 0.1402, 0.6700, 0.4000, 0.5366, 0.1030, 0.2620
0.6200, 1.0000, 0.3330, 0.1140, 0.1820, 0.1140, 0.3800, 0.5000, 0.5011, 0.0960, 0.2750
0.5700, 1.0000, 0.2560, 0.0960, 0.2000, 0.1330, 0.5200, 0.3850, 0.4318, 0.1050, 0.1770
0.2000, 1.0000, 0.2420, 0.0880, 0.1260, 0.0722, 0.4500, 0.3000, 0.3784, 0.0740, 0.0710
0.5300, 1.0000, 0.2210, 0.0980, 0.1650, 0.1052, 0.4700, 0.4000, 0.4159, 0.0810, 0.0470
0.3200, 1.0000, 0.3140, 0.0890, 0.1530, 0.0842, 0.5600, 0.3000, 0.4159, 0.0900, 0.1870
0.4100, 0.0000, 0.2310, 0.0860, 0.1480, 0.0780, 0.5800, 0.3000, 0.4094, 0.0600, 0.1250
0.6000, 0.0000, 0.2340, 0.0767, 0.2470, 0.1480, 0.6500, 0.3800, 0.5136, 0.0770, 0.0780
0.2600, 0.0000, 0.1880, 0.0830, 0.1910, 0.1036, 0.6900, 0.3000, 0.4522, 0.0690, 0.0510
0.3700, 0.0000, 0.3080, 0.1120, 0.2820, 0.1972, 0.4300, 0.7000, 0.5342, 0.1010, 0.2580
0.4500, 0.0000, 0.3200, 0.1100, 0.2240, 0.1342, 0.4500, 0.5000, 0.5412, 0.0930, 0.2150
0.6700, 0.0000, 0.3160, 0.1160, 0.1790, 0.0904, 0.4100, 0.4000, 0.5472, 0.1000, 0.3030
0.3400, 1.0000, 0.3550, 0.1200, 0.2330, 0.1466, 0.3400, 0.7000, 0.5568, 0.1010, 0.2430
0.5000, 0.0000, 0.3190, 0.0783, 0.2070, 0.1492, 0.3800, 0.5450, 0.4595, 0.0840, 0.0910
0.7100, 0.0000, 0.2950, 0.0970, 0.2270, 0.1516, 0.4500, 0.5000, 0.5024, 0.1080, 0.1500
0.5700, 1.0000, 0.3160, 0.1170, 0.2250, 0.1076, 0.4000, 0.6000, 0.5958, 0.1130, 0.3100
0.4900, 0.0000, 0.2030, 0.0930, 0.1840, 0.1030, 0.6100, 0.3000, 0.4605, 0.0930, 0.1530
0.3500, 0.0000, 0.4130, 0.0810, 0.1680, 0.1028, 0.3700, 0.5000, 0.4949, 0.0940, 0.3460
0.4100, 1.0000, 0.2120, 0.1020, 0.1840, 0.1004, 0.6400, 0.3000, 0.4585, 0.0790, 0.0630
0.7000, 1.0000, 0.2410, 0.0823, 0.1940, 0.1492, 0.3100, 0.6260, 0.4234, 0.1050, 0.0890
0.5200, 0.0000, 0.2300, 0.1070, 0.1790, 0.1237, 0.4250, 0.4210, 0.4159, 0.0930, 0.0500
0.6000, 0.0000, 0.2560, 0.0780, 0.1950, 0.0954, 0.9100, 0.2000, 0.3761, 0.0870, 0.0390
0.6200, 0.0000, 0.2250, 0.1250, 0.2150, 0.0990, 0.9800, 0.2000, 0.4500, 0.0950, 0.1030
0.4400, 1.0000, 0.3820, 0.1230, 0.2010, 0.1266, 0.4400, 0.5000, 0.5024, 0.0920, 0.3080
0.2800, 1.0000, 0.1920, 0.0810, 0.1550, 0.0946, 0.5100, 0.3000, 0.3850, 0.0870, 0.1160
0.5800, 1.0000, 0.2900, 0.0850, 0.1560, 0.1092, 0.3600, 0.4000, 0.3989, 0.0860, 0.1450
0.3900, 1.0000, 0.2400, 0.0897, 0.1900, 0.1136, 0.5200, 0.3650, 0.4804, 0.1010, 0.0740
0.3400, 1.0000, 0.2060, 0.0980, 0.1830, 0.0920, 0.8300, 0.2000, 0.3689, 0.0920, 0.0450
0.6500, 0.0000, 0.2630, 0.0700, 0.2440, 0.1662, 0.5100, 0.5000, 0.4898, 0.0980, 0.1150
0.6600, 1.0000, 0.3460, 0.1150, 0.2040, 0.1394, 0.3600, 0.6000, 0.4963, 0.1090, 0.2640
0.5100, 0.0000, 0.2340, 0.0870, 0.2200, 0.1088, 0.9300, 0.2000, 0.4511, 0.0820, 0.0870
0.5000, 1.0000, 0.2920, 0.1190, 0.1620, 0.0852, 0.5400, 0.3000, 0.4736, 0.0950, 0.2020
0.5900, 1.0000, 0.2720, 0.1070, 0.1580, 0.1020, 0.3900, 0.4000, 0.4443, 0.0930, 0.1270
0.5200, 0.0000, 0.2700, 0.0783, 0.1340, 0.0730, 0.4400, 0.3050, 0.4443, 0.0690, 0.1820
0.6900, 1.0000, 0.2450, 0.1080, 0.2430, 0.1364, 0.4000, 0.6000, 0.5808, 0.1000, 0.2410
0.5300, 0.0000, 0.2410, 0.1050, 0.1840, 0.1134, 0.4600, 0.4000, 0.4812, 0.0950, 0.0660
0.4700, 1.0000, 0.2530, 0.0980, 0.1730, 0.1056, 0.4400, 0.4000, 0.4762, 0.1080, 0.0940
0.5200, 0.0000, 0.2880, 0.1130, 0.2800, 0.1740, 0.6700, 0.4000, 0.5273, 0.0860, 0.2830
0.3900, 0.0000, 0.2090, 0.0950, 0.1500, 0.0656, 0.6800, 0.2000, 0.4407, 0.0950, 0.0640
0.6700, 1.0000, 0.2300, 0.0700, 0.1840, 0.1280, 0.3500, 0.5000, 0.4654, 0.0990, 0.1020
0.5900, 1.0000, 0.2410, 0.0960, 0.1700, 0.0986, 0.5400, 0.3000, 0.4466, 0.0850, 0.2000
0.5100, 1.0000, 0.2810, 0.1060, 0.2020, 0.1222, 0.5500, 0.4000, 0.4820, 0.0870, 0.2650
0.2300, 1.0000, 0.1800, 0.0780, 0.1710, 0.0960, 0.4800, 0.4000, 0.4905, 0.0920, 0.0940
0.6800, 0.0000, 0.2590, 0.0930, 0.2530, 0.1812, 0.5300, 0.5000, 0.4543, 0.0980, 0.2300
0.4400, 0.0000, 0.2150, 0.0850, 0.1570, 0.0922, 0.5500, 0.3000, 0.3892, 0.0840, 0.1810
0.6000, 1.0000, 0.2430, 0.1030, 0.1410, 0.0866, 0.3300, 0.4000, 0.4673, 0.0780, 0.1560
0.5200, 0.0000, 0.2450, 0.0900, 0.1980, 0.1290, 0.2900, 0.7000, 0.5298, 0.0860, 0.2330
0.3800, 0.0000, 0.2130, 0.0720, 0.1650, 0.0602, 0.8800, 0.2000, 0.4431, 0.0900, 0.0600
0.6100, 0.0000, 0.2580, 0.0900, 0.2800, 0.1954, 0.5500, 0.5000, 0.4997, 0.0900, 0.2190
0.6800, 1.0000, 0.2480, 0.1010, 0.2210, 0.1514, 0.6000, 0.4000, 0.3871, 0.0870, 0.0800
0.2800, 1.0000, 0.3150, 0.0830, 0.2280, 0.1494, 0.3800, 0.6000, 0.5313, 0.0830, 0.0680
0.6500, 1.0000, 0.3350, 0.1020, 0.1900, 0.1262, 0.3500, 0.5000, 0.4970, 0.1020, 0.3320
0.6900, 0.0000, 0.2810, 0.1130, 0.2340, 0.1428, 0.5200, 0.4000, 0.5278, 0.0770, 0.2480
0.5100, 0.0000, 0.2430, 0.0853, 0.1530, 0.0716, 0.7100, 0.2150, 0.3951, 0.0820, 0.0840
0.2900, 0.0000, 0.3500, 0.0983, 0.2040, 0.1426, 0.5000, 0.4080, 0.4043, 0.0910, 0.2000
0.5500, 1.0000, 0.2350, 0.0930, 0.1770, 0.1268, 0.4100, 0.4000, 0.3829, 0.0830, 0.0550
0.3400, 1.0000, 0.3000, 0.0830, 0.1850, 0.1072, 0.5300, 0.3000, 0.4820, 0.0920, 0.0850
0.6700, 0.0000, 0.2070, 0.0830, 0.1700, 0.0998, 0.5900, 0.3000, 0.4025, 0.0770, 0.0890
0.4900, 0.0000, 0.2560, 0.0760, 0.1610, 0.0998, 0.5100, 0.3000, 0.3932, 0.0780, 0.0310
0.5500, 1.0000, 0.2290, 0.0810, 0.1230, 0.0672, 0.4100, 0.3000, 0.4304, 0.0880, 0.1290
0.5900, 1.0000, 0.2510, 0.0900, 0.1630, 0.1014, 0.4600, 0.4000, 0.4357, 0.0910, 0.0830
0.5300, 0.0000, 0.3320, 0.0827, 0.1860, 0.1068, 0.4600, 0.4040, 0.5112, 0.1020, 0.2750
0.4800, 1.0000, 0.2410, 0.1100, 0.2090, 0.1346, 0.5800, 0.4000, 0.4407, 0.1000, 0.0650
0.5200, 0.0000, 0.2950, 0.1043, 0.2110, 0.1328, 0.4900, 0.4310, 0.4984, 0.0980, 0.1980
0.6900, 0.0000, 0.2960, 0.1220, 0.2310, 0.1284, 0.5600, 0.4000, 0.5451, 0.0860, 0.2360
0.6000, 1.0000, 0.2280, 0.1100, 0.2450, 0.1898, 0.3900, 0.6000, 0.4394, 0.0880, 0.2530
0.4600, 1.0000, 0.2270, 0.0830, 0.1830, 0.1258, 0.3200, 0.6000, 0.4836, 0.0750, 0.1240
0.5100, 1.0000, 0.2620, 0.1010, 0.1610, 0.0996, 0.4800, 0.3000, 0.4205, 0.0880, 0.0440
0.6700, 1.0000, 0.2350, 0.0960, 0.2070, 0.1382, 0.4200, 0.5000, 0.4898, 0.1110, 0.1720
0.4900, 0.0000, 0.2210, 0.0850, 0.1360, 0.0634, 0.6200, 0.2190, 0.3970, 0.0720, 0.1140
0.4600, 1.0000, 0.2650, 0.0940, 0.2470, 0.1602, 0.5900, 0.4000, 0.4935, 0.1110, 0.1420
0.4700, 0.0000, 0.3240, 0.1050, 0.1880, 0.1250, 0.4600, 0.4090, 0.4443, 0.0990, 0.1090
0.7500, 0.0000, 0.3010, 0.0780, 0.2220, 0.1542, 0.4400, 0.5050, 0.4779, 0.0970, 0.1800
0.2800, 0.0000, 0.2420, 0.0930, 0.1740, 0.1064, 0.5400, 0.3000, 0.4220, 0.0840, 0.1440
0.6500, 1.0000, 0.3130, 0.1100, 0.2130, 0.1280, 0.4700, 0.5000, 0.5247, 0.0910, 0.1630
0.4200, 0.0000, 0.3010, 0.0910, 0.1820, 0.1148, 0.4900, 0.4000, 0.4511, 0.0820, 0.1470
0.5100, 0.0000, 0.2450, 0.0790, 0.2120, 0.1286, 0.6500, 0.3000, 0.4522, 0.0910, 0.0970
0.5300, 1.0000, 0.2770, 0.0950, 0.1900, 0.1018, 0.4100, 0.5000, 0.5464, 0.1010, 0.2200
0.5400, 0.0000, 0.2320, 0.1107, 0.2380, 0.1628, 0.4800, 0.4960, 0.4913, 0.1080, 0.1900
0.7300, 0.0000, 0.2700, 0.1020, 0.2110, 0.1210, 0.6700, 0.3000, 0.4745, 0.0990, 0.1090
0.5400, 0.0000, 0.2680, 0.1080, 0.1760, 0.0806, 0.6700, 0.3000, 0.4956, 0.1060, 0.1910
0.4200, 0.0000, 0.2920, 0.0930, 0.2490, 0.1742, 0.4500, 0.6000, 0.5004, 0.0920, 0.1220
0.7500, 0.0000, 0.3120, 0.1177, 0.2290, 0.1388, 0.2900, 0.7900, 0.5724, 0.1060, 0.2300
0.5500, 1.0000, 0.3210, 0.1127, 0.2070, 0.0924, 0.2500, 0.8280, 0.6105, 0.1110, 0.2420
0.6800, 1.0000, 0.2570, 0.1090, 0.2330, 0.1126, 0.3500, 0.7000, 0.6057, 0.1050, 0.2480
0.5700, 0.0000, 0.2690, 0.0980, 0.2460, 0.1652, 0.3800, 0.7000, 0.5366, 0.0960, 0.2490
0.4800, 0.0000, 0.3140, 0.0753, 0.2420, 0.1516, 0.3800, 0.6370, 0.5568, 0.1030, 0.1920
0.6100, 1.0000, 0.2560, 0.0850, 0.1840, 0.1162, 0.3900, 0.5000, 0.4970, 0.0980, 0.1310
0.6900, 0.0000, 0.3700, 0.1030, 0.2070, 0.1314, 0.5500, 0.4000, 0.4635, 0.0900, 0.2370
0.3800, 0.0000, 0.3260, 0.0770, 0.1680, 0.1006, 0.4700, 0.4000, 0.4625, 0.0960, 0.0780
0.4500, 1.0000, 0.2120, 0.0940, 0.1690, 0.0968, 0.5500, 0.3000, 0.4454, 0.1020, 0.1350
0.5100, 1.0000, 0.2920, 0.1070, 0.1870, 0.1390, 0.3200, 0.6000, 0.4382, 0.0950, 0.2440
0.7100, 1.0000, 0.2400, 0.0840, 0.1380, 0.0858, 0.3900, 0.4000, 0.4190, 0.0900, 0.1990
0.5700, 0.0000, 0.3610, 0.1170, 0.1810, 0.1082, 0.3400, 0.5000, 0.5268, 0.1000, 0.2700
0.5600, 1.0000, 0.2580, 0.1030, 0.1770, 0.1144, 0.3400, 0.5000, 0.4963, 0.0990, 0.1640
0.3200, 1.0000, 0.2200, 0.0880, 0.1370, 0.0786, 0.4800, 0.3000, 0.3951, 0.0780, 0.0720
0.5000, 0.0000, 0.2190, 0.0910, 0.1900, 0.1112, 0.6700, 0.3000, 0.4078, 0.0770, 0.0960
0.4300, 0.0000, 0.3430, 0.0840, 0.2560, 0.1726, 0.3300, 0.8000, 0.5529, 0.1040, 0.3060
0.5400, 1.0000, 0.2520, 0.1150, 0.1810, 0.1200, 0.3900, 0.5000, 0.4701, 0.0920, 0.0910
0.3100, 0.0000, 0.2330, 0.0850, 0.1900, 0.1308, 0.4300, 0.4000, 0.4394, 0.0770, 0.2140
0.5600, 0.0000, 0.2570, 0.0800, 0.2440, 0.1516, 0.5900, 0.4000, 0.5118, 0.0950, 0.0950
0.4400, 0.0000, 0.2510, 0.1330, 0.1820, 0.1130, 0.5500, 0.3000, 0.4249, 0.0840, 0.2160
0.5700, 1.0000, 0.3190, 0.1110, 0.1730, 0.1162, 0.4100, 0.4000, 0.4369, 0.0870, 0.2630

Test data:


# diabetes_norm_test_100.txt
#
0.6400, 1.0000, 0.2840, 0.1110, 0.1840, 0.1270, 0.4100, 0.4000, 0.4382, 0.0970, 0.1780
0.4300, 0.0000, 0.2810, 0.1210, 0.1920, 0.1210, 0.6000, 0.3000, 0.4007, 0.0930, 0.1130
0.1900, 0.0000, 0.2530, 0.0830, 0.2250, 0.1566, 0.4600, 0.5000, 0.4719, 0.0840, 0.2000
0.7100, 1.0000, 0.2610, 0.0850, 0.2200, 0.1524, 0.4700, 0.5000, 0.4635, 0.0910, 0.1390
0.5000, 1.0000, 0.2800, 0.1040, 0.2820, 0.1968, 0.4400, 0.6000, 0.5328, 0.0950, 0.1390
0.5900, 1.0000, 0.2360, 0.0730, 0.1800, 0.1074, 0.5100, 0.4000, 0.4682, 0.0840, 0.0880
0.5700, 0.0000, 0.2450, 0.0930, 0.1860, 0.0966, 0.7100, 0.3000, 0.4522, 0.0910, 0.1480
0.4900, 1.0000, 0.2100, 0.0820, 0.1190, 0.0854, 0.2300, 0.5000, 0.3970, 0.0740, 0.0880
0.4100, 1.0000, 0.3200, 0.1260, 0.1980, 0.1042, 0.4900, 0.4000, 0.5412, 0.1240, 0.2430
0.2500, 1.0000, 0.2260, 0.0850, 0.1300, 0.0710, 0.4800, 0.3000, 0.4007, 0.0810, 0.0710
0.5200, 1.0000, 0.1970, 0.0810, 0.1520, 0.0534, 0.8200, 0.2000, 0.4419, 0.0820, 0.0770
0.3400, 0.0000, 0.2120, 0.0840, 0.2540, 0.1134, 0.5200, 0.5000, 0.6094, 0.0920, 0.1090
0.4200, 1.0000, 0.3060, 0.1010, 0.2690, 0.1722, 0.5000, 0.5000, 0.5455, 0.1060, 0.2720
0.2800, 1.0000, 0.2550, 0.0990, 0.1620, 0.1016, 0.4600, 0.4000, 0.4277, 0.0940, 0.0600
0.4700, 1.0000, 0.2330, 0.0900, 0.1950, 0.1258, 0.5400, 0.4000, 0.4331, 0.0730, 0.0540
0.3200, 1.0000, 0.3100, 0.1000, 0.1770, 0.0962, 0.4500, 0.4000, 0.5187, 0.0770, 0.2210
0.4300, 0.0000, 0.1850, 0.0870, 0.1630, 0.0936, 0.6100, 0.2670, 0.3738, 0.0800, 0.0900
0.5900, 1.0000, 0.2690, 0.1040, 0.1940, 0.1266, 0.4300, 0.5000, 0.4804, 0.1060, 0.3110
0.5300, 0.0000, 0.2830, 0.1010, 0.1790, 0.1070, 0.4800, 0.4000, 0.4788, 0.1010, 0.2810
0.6000, 0.0000, 0.2570, 0.1030, 0.1580, 0.0846, 0.6400, 0.2000, 0.3850, 0.0970, 0.1820
0.5400, 1.0000, 0.3610, 0.1150, 0.1630, 0.0984, 0.4300, 0.4000, 0.4682, 0.1010, 0.3210
0.3500, 1.0000, 0.2410, 0.0947, 0.1550, 0.0974, 0.3200, 0.4840, 0.4852, 0.0940, 0.0580
0.4900, 1.0000, 0.2580, 0.0890, 0.1820, 0.1186, 0.3900, 0.5000, 0.4804, 0.1150, 0.2620
0.5800, 0.0000, 0.2280, 0.0910, 0.1960, 0.1188, 0.4800, 0.4000, 0.4984, 0.1150, 0.2060
0.3600, 1.0000, 0.3910, 0.0900, 0.2190, 0.1358, 0.3800, 0.6000, 0.5421, 0.1030, 0.2330
0.4600, 1.0000, 0.4220, 0.0990, 0.2110, 0.1370, 0.4400, 0.5000, 0.5011, 0.0990, 0.2420
0.4400, 1.0000, 0.2660, 0.0990, 0.2050, 0.1090, 0.4300, 0.5000, 0.5580, 0.1110, 0.1230
0.4600, 0.0000, 0.2990, 0.0830, 0.1710, 0.1130, 0.3800, 0.4500, 0.4585, 0.0980, 0.1670
0.5400, 0.0000, 0.2100, 0.0780, 0.1880, 0.1074, 0.7000, 0.3000, 0.3970, 0.0730, 0.0630
0.6300, 1.0000, 0.2550, 0.1090, 0.2260, 0.1032, 0.4600, 0.5000, 0.5951, 0.0870, 0.1970
0.4100, 1.0000, 0.2420, 0.0900, 0.1990, 0.1236, 0.5700, 0.4000, 0.4522, 0.0860, 0.0710
0.2800, 0.0000, 0.2540, 0.0930, 0.1410, 0.0790, 0.4900, 0.3000, 0.4174, 0.0910, 0.1680
0.1900, 0.0000, 0.2320, 0.0750, 0.1430, 0.0704, 0.5200, 0.3000, 0.4635, 0.0720, 0.1400
0.6100, 1.0000, 0.2610, 0.1260, 0.2150, 0.1298, 0.5700, 0.4000, 0.4949, 0.0960, 0.2170
0.4800, 0.0000, 0.3270, 0.0930, 0.2760, 0.1986, 0.4300, 0.6420, 0.5148, 0.0910, 0.1210
0.5400, 1.0000, 0.2730, 0.1000, 0.2000, 0.1440, 0.3300, 0.6000, 0.4745, 0.0760, 0.2350
0.5300, 1.0000, 0.2660, 0.0930, 0.1850, 0.1224, 0.3600, 0.5000, 0.4890, 0.0820, 0.2450
0.4800, 0.0000, 0.2280, 0.1010, 0.1100, 0.0416, 0.5600, 0.2000, 0.4127, 0.0970, 0.0400
0.5300, 0.0000, 0.2880, 0.1117, 0.1450, 0.0872, 0.4600, 0.3150, 0.4078, 0.0850, 0.0520
0.2900, 1.0000, 0.1810, 0.0730, 0.1580, 0.0990, 0.4100, 0.4000, 0.4500, 0.0780, 0.1040
0.6200, 0.0000, 0.3200, 0.0880, 0.1720, 0.0690, 0.3800, 0.4000, 0.5784, 0.1000, 0.1320
0.5000, 1.0000, 0.2370, 0.0920, 0.1660, 0.0970, 0.5200, 0.3000, 0.4443, 0.0930, 0.0880
0.5800, 1.0000, 0.2360, 0.0960, 0.2570, 0.1710, 0.5900, 0.4000, 0.4905, 0.0820, 0.0690
0.5500, 1.0000, 0.2460, 0.1090, 0.1430, 0.0764, 0.5100, 0.3000, 0.4357, 0.0880, 0.2190
0.5400, 0.0000, 0.2260, 0.0900, 0.1830, 0.1042, 0.6400, 0.3000, 0.4304, 0.0920, 0.0720
0.3600, 0.0000, 0.2780, 0.0730, 0.1530, 0.1044, 0.4200, 0.4000, 0.3497, 0.0730, 0.2010
0.6300, 1.0000, 0.2410, 0.1110, 0.1840, 0.1122, 0.4400, 0.4000, 0.4935, 0.0820, 0.1100
0.4700, 1.0000, 0.2650, 0.0700, 0.1810, 0.1048, 0.6300, 0.3000, 0.4190, 0.0700, 0.0510
0.5100, 1.0000, 0.3280, 0.1120, 0.2020, 0.1006, 0.3700, 0.5000, 0.5775, 0.1090, 0.2770
0.4200, 0.0000, 0.1990, 0.0760, 0.1460, 0.0832, 0.5500, 0.3000, 0.3664, 0.0790, 0.0630
0.3700, 1.0000, 0.2360, 0.0940, 0.2050, 0.1388, 0.5300, 0.4000, 0.4190, 0.1070, 0.1180
0.2800, 0.0000, 0.2210, 0.0820, 0.1680, 0.1006, 0.5400, 0.3000, 0.4205, 0.0860, 0.0690
0.5800, 0.0000, 0.2810, 0.1110, 0.1980, 0.0806, 0.3100, 0.6000, 0.6068, 0.0930, 0.2730
0.3200, 0.0000, 0.2650, 0.0860, 0.1840, 0.1016, 0.5300, 0.4000, 0.4990, 0.0780, 0.2580
0.2500, 1.0000, 0.2350, 0.0880, 0.1430, 0.0808, 0.5500, 0.3000, 0.3584, 0.0830, 0.0430
0.6300, 0.0000, 0.2600, 0.0857, 0.1550, 0.0782, 0.4600, 0.3370, 0.5037, 0.0970, 0.1980
0.5200, 0.0000, 0.2780, 0.0850, 0.2190, 0.1360, 0.4900, 0.4000, 0.5136, 0.0750, 0.2420
0.6500, 1.0000, 0.2850, 0.1090, 0.2010, 0.1230, 0.4600, 0.4000, 0.5075, 0.0960, 0.2320
0.4200, 0.0000, 0.3060, 0.1210, 0.1760, 0.0928, 0.6900, 0.3000, 0.4263, 0.0890, 0.1750
0.5300, 0.0000, 0.2220, 0.0780, 0.1640, 0.0810, 0.7000, 0.2000, 0.4174, 0.1010, 0.0930
0.7900, 1.0000, 0.2330, 0.0880, 0.1860, 0.1284, 0.3300, 0.6000, 0.4812, 0.1020, 0.1680
0.4300, 0.0000, 0.3540, 0.0930, 0.1850, 0.1002, 0.4400, 0.4000, 0.5318, 0.1010, 0.2750
0.4400, 0.0000, 0.3140, 0.1150, 0.1650, 0.0976, 0.5200, 0.3000, 0.4344, 0.0890, 0.2930
0.6200, 1.0000, 0.3780, 0.1190, 0.1130, 0.0510, 0.3100, 0.4000, 0.5043, 0.0840, 0.2810
0.3300, 0.0000, 0.1890, 0.0700, 0.1620, 0.0918, 0.5900, 0.3000, 0.4025, 0.0580, 0.0720
0.5600, 0.0000, 0.3500, 0.0793, 0.1950, 0.1408, 0.4200, 0.4640, 0.4111, 0.0960, 0.1400
0.6600, 0.0000, 0.2170, 0.1260, 0.2120, 0.1278, 0.4500, 0.4710, 0.5278, 0.1010, 0.1890
0.3400, 1.0000, 0.2530, 0.1110, 0.2300, 0.1620, 0.3900, 0.6000, 0.4977, 0.0900, 0.1810
0.4600, 1.0000, 0.2380, 0.0970, 0.2240, 0.1392, 0.4200, 0.5000, 0.5366, 0.0810, 0.2090
0.5000, 0.0000, 0.3180, 0.0820, 0.1360, 0.0692, 0.5500, 0.2000, 0.4078, 0.0850, 0.1360
0.6900, 0.0000, 0.3430, 0.1130, 0.2000, 0.1238, 0.5400, 0.4000, 0.4710, 0.1120, 0.2610
0.3400, 0.0000, 0.2630, 0.0870, 0.1970, 0.1200, 0.6300, 0.3000, 0.4249, 0.0960, 0.1130
0.7100, 1.0000, 0.2700, 0.0933, 0.2690, 0.1902, 0.4100, 0.6560, 0.5242, 0.0930, 0.1310
0.4700, 0.0000, 0.2720, 0.0800, 0.2080, 0.1456, 0.3800, 0.6000, 0.4804, 0.0920, 0.1740
0.4100, 0.0000, 0.3380, 0.1233, 0.1870, 0.1270, 0.4500, 0.4160, 0.4318, 0.1000, 0.2570
0.3400, 0.0000, 0.3300, 0.0730, 0.1780, 0.1146, 0.5100, 0.3490, 0.4127, 0.0920, 0.0550
0.5100, 0.0000, 0.2410, 0.0870, 0.2610, 0.1756, 0.6900, 0.4000, 0.4407, 0.0930, 0.0840
0.4300, 0.0000, 0.2130, 0.0790, 0.1410, 0.0788, 0.5300, 0.3000, 0.3829, 0.0900, 0.0420
0.5500, 0.0000, 0.2300, 0.0947, 0.1900, 0.1376, 0.3800, 0.5000, 0.4277, 0.1060, 0.1460
0.5900, 1.0000, 0.2790, 0.1010, 0.2180, 0.1442, 0.3800, 0.6000, 0.5187, 0.0950, 0.2120
0.2700, 1.0000, 0.3360, 0.1100, 0.2460, 0.1566, 0.5700, 0.4000, 0.5088, 0.0890, 0.2330
0.5100, 1.0000, 0.2270, 0.1030, 0.2170, 0.1624, 0.3000, 0.7000, 0.4812, 0.0800, 0.0910
0.4900, 1.0000, 0.2740, 0.0890, 0.1770, 0.1130, 0.3700, 0.5000, 0.4905, 0.0970, 0.1110
0.2700, 0.0000, 0.2260, 0.0710, 0.1160, 0.0434, 0.5600, 0.2000, 0.4419, 0.0790, 0.1520
0.5700, 1.0000, 0.2320, 0.1073, 0.2310, 0.1594, 0.4100, 0.5630, 0.5030, 0.1120, 0.1200
0.3900, 1.0000, 0.2690, 0.0930, 0.1360, 0.0754, 0.4800, 0.3000, 0.4143, 0.0990, 0.0670
0.6200, 1.0000, 0.3460, 0.1200, 0.2150, 0.1292, 0.4300, 0.5000, 0.5366, 0.1230, 0.3100
0.3700, 0.0000, 0.2330, 0.0880, 0.2230, 0.1420, 0.6500, 0.3400, 0.4357, 0.0820, 0.0940
0.4600, 0.0000, 0.2110, 0.0800, 0.2050, 0.1444, 0.4200, 0.5000, 0.4533, 0.0870, 0.1830
0.6800, 1.0000, 0.2350, 0.1010, 0.1620, 0.0854, 0.5900, 0.3000, 0.4477, 0.0910, 0.0660
0.5100, 0.0000, 0.3150, 0.0930, 0.2310, 0.1440, 0.4900, 0.4700, 0.5252, 0.1170, 0.1730
0.4100, 0.0000, 0.2080, 0.0860, 0.2230, 0.1282, 0.8300, 0.3000, 0.4078, 0.0890, 0.0720
0.5300, 0.0000, 0.2650, 0.0970, 0.1930, 0.1224, 0.5800, 0.3000, 0.4143, 0.0990, 0.0490
0.4500, 0.0000, 0.2420, 0.0830, 0.1770, 0.1184, 0.4500, 0.4000, 0.4220, 0.0820, 0.0640
0.3300, 0.0000, 0.1950, 0.0800, 0.1710, 0.0854, 0.7500, 0.2000, 0.3970, 0.0800, 0.0480
0.6000, 1.0000, 0.2820, 0.1120, 0.1850, 0.1138, 0.4200, 0.4000, 0.4984, 0.0930, 0.1780
0.4700, 1.0000, 0.2490, 0.0750, 0.2250, 0.1660, 0.4200, 0.5000, 0.4443, 0.1020, 0.1040
0.6000, 1.0000, 0.2490, 0.0997, 0.1620, 0.1066, 0.4300, 0.3770, 0.4127, 0.0950, 0.1320
0.3600, 0.0000, 0.3000, 0.0950, 0.2010, 0.1252, 0.4200, 0.4790, 0.5130, 0.0850, 0.2200
0.3600, 0.0000, 0.1960, 0.0710, 0.2500, 0.1332, 0.9700, 0.3000, 0.4595, 0.0920, 0.0570
Posted in Machine Learning, Scikit | Leave a comment

Comparing Support Vector Regression versus Post-Training-Trimmed Kernel Ridge Regression Using Scikit

Posted on June 9, 2026 by jamesdmccaffrey

Bottom line: I did a short experiment to compare support vector regression (SVR) with post-training-trimmed kernel ridge regression (KRR). I used the scikit library SVR and KernelRidge modules. Because of the nearly infinite number of hyperparameter values involved, it wasn’t possible to draw a strong conclusion, but my results strongly suggest the two techniques are essentially the same, or at least very similar, from a practical point of view.

Bear with me. The goal of a machine learning regression problem is to predict a single numeric value. For example, a bank might want to predict a maximum loan amount based on applicant age, sex, annual income, debt, and so on.

Common regression techniques include linear regression, quadratic regression, nearest neighbors regression, kernel ridge regression, Gaussian process regression, kernel support vector regression, neural network regression, random forest regression, and gradient boost regression. Each technique has dozens of variations, and each technique has pros and cons.

Two closely related machine learning regression techniques are kernel ridge regression (KRR) and support vector regression (SVR). Both techniques use a kernel function (usually the radial basis function, RBF) to compare two data items for similarity. Both techniques must store training data in order to make predictions. KRR must store all training items, while SVR eliminates some of the items during training, leaving just the “support vectors” that need to be stored. So, SVR uses less memory than KRR.

On the other hand, the loss function used by SVR is not differentiable, so SVR cannot be trained using stochastic gradient descent, and therefore, SVR does not scale well to very large datasets. So, KRR is easier to train than SVR.

Note: In addition to the kernel support vector regression technique discussed in this blog post, there is a linear support vector regression technique which is essentially useless in practice.

The idea of combining KRR and SVR goes like this: Briefly, I train a KRR model as normal using all data. Then I identify training data items that are predicted “too well” and remove them, leaving just pseudo support vectors. The somewhat unobvious SVR idea is that items that are predicted too well don’t help model accuracy very much, and can lead to an overfitted model. After removing some training items, I retrain a new KRR model using only the reduced training data. This gives a trimmed/sparse KRR model that approximates an SVR model.

In my mind, this technique gives the advantages of KRR (ability to handle very large datasets via SGD training) and the advantages of SVR (fewer stored model items and weights than KRR).

I first implemented the idea using the scikit KernelRidge module and the idea seemed to work well. But I wanted to take a second look, using a very small dataset, as a sanity check to make sure there weren’t any unexpected, bad, surprises. The output of one of the experiments is:

Begin SVR vs. trimmed KRR demo

Creating 20-item, 5-feature train data

X:
[[ 0.0458 -0.1872  1.4694 -1.4544  1.5328]
 [ 0.9008  0.4657  1.4883 -1.1651 -1.5362]
 [ 0.6536  0.8644  2.2698 -2.5530 -0.7422]
 [ 0.4002  0.9787  1.8676  1.7641  2.2409]
 [ 0.3869 -0.5108 -0.0282 -0.8955 -1.1806]
 [ 1.2303  1.2024 -0.3023  0.1563 -0.3873]
 [ 1.4543  0.7610  0.4439  0.1440  0.1217]
 [ 1.4941 -0.2052 -0.8541  0.3337  0.3131]
 [-0.3596 -0.8131  0.1774 -0.6725 -1.7263]
 [ 1.1788 -0.1799  1.0545  1.8959 -1.0708]
 [ 0.1290  1.1394  0.4023  0.7291 -1.2348]
 [ 0.9501 -0.1514  0.4106 -0.9773 -0.1032]
 [ 1.2224  0.2083  0.3564 -0.4032  0.9766]
 [-1.6302  0.4628  0.0519 -0.4018 -0.9073]
 [ 0.3782 -0.8878 -0.3479  0.1549 -1.9808]
 [ 0.0665  0.3025 -0.3627  0.4283 -0.6343]
 [ 0.0105  1.7859  0.4020  0.7066  0.1269]
 [-1.4200 -1.7063 -0.5097 -1.0486  1.9508]
 [-1.2528  0.7775 -0.2127 -0.4381 -1.6139]
 [-0.8708 -0.5788  0.0562 -0.6848 -0.3116]]

Creating and training SVR model
gamma = 0.1000
C = 1000.0000
epsilon = 8.0000
Done

Number support vectors: 10

Support vector indices:
[ 0  1  3  9 10 14 16 17 18 19]

Accuracy (within 0.10) train = 0.6000
MSE train = 64.0016
R2 train = 0.996314

===============================

Creating and training preliminary KRR RBF model
Setting gamma = 0.1000
Setting alpha = 0.0010
Done

Removing non-pseudo-support vectors
Using KRR trim epsilon = 0.1000

Number of pseudo support vectors = 10

Support vector indices:
[ 1  3  9 10 11 15 16 17 18 19]

Re-training trimmed KRR model
Done

Accuracy (within 0.10) train = 1.0000
MSE train = 0.0225
R2 train = 0.999999

End demo

The demo program begins by using the scikit make_regression() function to create a tiny synthetic dataset with just 20 rows/items, each with 5 predictor values. An SVR model ended up with 10 support vectors and an R2 score (essentially a scaled accuracy metric) of 0.9963 on the training data.

I tuned a trimmed KRR model to one with 10 pseudo-support vectors. The two techniques had 8 out of 10 support vectors in common — that’s good.

The trimmed KRR model had an R2 score of 0.9999 on the training data. The trimmed KRR model had better MSE than the SVR model, but that was expected because KRR minimizes MSE and SVR does not. The much better accuracy of the trimmed KRR model (100% for KRR vs. 60% for SVR) isn’t significant because the dataset is so small, and I didn’t use a test dataset to measure possible overfitting.

I noticed that both models were ultra-sensitive to hyperparameter values — gamma, C, and epsilon for SVR, and gamma, alpha, and trim-epsilon for trimmed KRR. This is a major weakness of both techniques.

Anyway, the sanity check was successful. It is possible to approximate a SVR regression model using a trimmed KRR model.


It’s common for me to refactor the software systems I create many times. It’s rarely possible to a get a non-trivial system completely correct on the first effort, and so a system is usually a collection of software sequels, so to speak, where each sequel is a bit better than its predecessor.

I’m a big fan of science fiction movies. There have been dozens of sci fi sequels. “Star Wars” (1977) and “Star Wars: The Empire Strikes Back” (1980). “Alien” (1979) and “Aliens” (1986). And so on. But most movie sequels are worse than their predecessor, however there are exceptions where the sequel is as-good-as, or even better than the original.

Left: In “The Quatermass Xperiment” aka “The Creeping Unknown” (1955), Dr. Bernard Quatermass leads a British effort to put men into space. Three astronauts are sent up, but only one man returns. He has been exposed to cosmic radiation and morphs into a deadly blob-like creature. The creature is eventually electrocuted. My personal grade = B+.

Right: In “Quatermass 2” aka “Enemy From Space” (1957), Quatermass discovers an alien invasion. The aliens use parasites to control the people in a small village and use the villagers to construct a plant to create alien food to support the invasion. Quatermass and the UK military blow up the plant and defeat the alien’s plans. My personal grade = A-.

Demo program. Replace the “lt” (less than) and “gt” with the Boolean operator symbols. (My blog editor chokes on symbols).

# svr_vs_krr_trimmed_scikit.py
# train using KRR, remove some, retrain using KRR

import numpy as np
from sklearn.kernel_ridge import KernelRidge
from sklearn.svm import SVR
from sklearn.datasets import make_regression

# KernelRidge(alpha=1, *, kernel='linear', gamma=None,
# degree=3, coef0=1, kernel_params=None)

# SVR(*, kernel='rbf', degree=3, gamma='scale',
# coef0=0.0, tol=0.001, C=1.0, epsilon=0.1,
# shrinking=True, cache_size=200, verbose=False,
# max_iter=-1)

# make_regression(n_samples=100, n_features=100,
# *, n_informative=10, n_targets=1, bias=0.0,
# effective_rank=None, tail_strength=0.5, noise=0.0,
# shuffle=True, coef=False, random_state=None)

# -----------------------------------------------------------

np.set_printoptions(precision=4, suppress=True,
  floatmode='fixed', linewidth=60)

# -----------------------------------------------------------

def accuracy(model, data_X, data_y, pct_close):
  n = len(data_X)
  n_correct = 0; n_wrong = 0
  for i in range(n):
    x = data_X[i].reshape(1,-1)
    y = data_y[i]
    y_pred = model.predict(x)[0]

    if np.abs(y - y_pred) "lt" np.abs(y * pct_close):
      n_correct += 1
    else: 
      n_wrong += 1
  return n_correct / (n_correct + n_wrong)

def mse(model, data_X, data_y):
  n = len(data_X)
  sum = 0.0
  for i in range(n):
    actual_y = data_y[i]
    pred_y = model.predict(data_X[i].reshape(1, -1))[0]
    diff = actual_y - pred_y
    sum += diff * diff
  return sum /n

# -----------------------------------------------------------
# -----------------------------------------------------------

print("\nBegin SVR vs. trimmmed KRR demo ")

np.set_printoptions(precision=4, suppress=True,
    floatmode='fixed')

print("\nCreating 20-item, 5-feature train data ")
X, y = make_regression(n_samples=20, n_features=5,
  n_informative=5, noise=0, random_state=0)

print("\nX: ")
print(X)

print("\nCreating and training SVR model ")
svr_gamma = 0.10
svr_C = 1000.0
svr_epsilon = 8.0
print("gamma = %0.4f " % svr_gamma)
print("C = %0.4f " % svr_C)
print("epsilon = %0.4f " % svr_epsilon)
svr_model = SVR(kernel='rbf', gamma=svr_gamma, C=svr_C,
  epsilon = svr_epsilon)
svr_model.fit(X, y)
print("Done ")

print("\nNumber support vectors: ", end="")
print(len(svr_model.support_))
print("\nSupport vector indices: ")
print(svr_model.support_)

X_sv = X[svr_model.support_,:]
y_sv = y[svr_model.support_]

acc_train = accuracy(svr_model, X_sv, y_sv, 0.10)
print("\nAccuracy (within 0.10) train = %0.4f " % \
  acc_train)
mse_train = mse(svr_model, X_sv, y_sv)
print("MSE train = %0.4f " % mse_train)
svr_r2 = svr_model.score(X_sv, y_sv)
print("R2 train = %0.6f " % svr_r2)

print("\n=============================== ")

print("\nCreating and training preliminary KRR model ")
krr_gamma = 0.1000
krr_alpha = 0.0010
print("Setting gamma = %0.4f " % krr_gamma)
print("Setting alpha = %0.4f " % krr_alpha)
model = KernelRidge(kernel='rbf', gamma=krr_gamma,
  alpha=krr_alpha)
model.fit(X, y)
print("Done ")

print("\nRemoving non-pseudo-support vectors ")
krr_trim_epsilon = 0.10
print("Using KRR trim epsilon = %0.4f " % krr_trim_epsilon)
# smaller epsilon = more support vectors
# larger epsilon = fewer support vectors

predictions = model.predict(X)
residuals = np.abs(y - predictions)
# keep only pts outside epsilon-tube (pseudo support vecs)
sv_indices = np.where(residuals "gt" krr_trim_epsilon)[0]
X_sv = X[sv_indices]
y_sv = y[sv_indices]
print("\nNumber of pseudo support vectors = " + \
  str(len(X_sv)))
print("\nSupport vector indices: ")
print(sv_indices)

# retrain
print("\nRe-training trimmed KRR model ")
krr_model = KernelRidge(kernel='rbf', 
  gamma=krr_gamma, alpha=krr_alpha)
krr_model.fit(X_sv, y_sv)
print("Done ")

acc_train = accuracy(krr_model, X_sv, y_sv, 0.10)
print("\nAccuracy (within 0.10) train = %0.4f " % \
  acc_train)
mse_train = mse(krr_model, X_sv, y_sv)
print("MSE train = %0.4f " % mse_train)
krr_r2 = krr_model.score(X_sv, y_sv)
print("R2 train = %0.6f " % krr_r2)

print("\nEnd demo ")
Posted in Machine Learning, Scikit | Leave a comment