I presented a talk titled “Neural Exotica: Beyond Basic Architectures” at the 2022 Predictive Analytics World conference. The event ran June 19-24 and was in Las Vegas.

The Predictive Analytics World (PAW) conference was co-located with the Machine Learning Week (MLW) conference. The events share the same conference space but have different areas of emphasis. PAW focuses on industry applications of all kinds of predictive systems, with an emphasis on practical systems. MLW focuses on more technical topics, mostly related to neural technologies and advanced classical ML techniques.

Left: My room before attendees arrived. Right: The conference had a small Expo.

In addition to my presentation at PAW, earlier in the week of the events, I presented a full-day workshop as part of the MLW events.

My talk had about 60 people in the audience. They asked excellent questions.

Two of the slides from my talk. Left: This slide indicates how difficult GANs are (and therefore the appeal of the new diffusion architecture). Right: A recap.

In my PAW talk, I described five neural architectures: autoencoders, variational autoencoders (VAEs), generative adversarial networks (GANs), Siamese networks, and transformer architecture networks. I also mentioned the new diffusion architecture networks, which may replace GANs for high resolution image generation.

I enjoyed both the PAW and MLW events a lot. The venue and logistics were excellent, and I gained a lot of valuable information from conversations with attendees.