Developing a community of MN AI enthusiasts.

Notice: Parking and entrance is on backside of building facing Hwy 94. Door is near far end of parking lot.

Agenda:
Networking - 7:00-7:15

Demonstration/Q&A: 7:15-8:05
Sergey: New UI paradigm = AI + Content + Context + Conversation. Using and authoring.

Tom: U-nets and generative adversarial networks (GANs) have remarkable parallels and stark differences when compared our brain's natural memory network in the hippocampus and cortex. Tom will briefly review U-nets, GANs, hippocampus, and neocortex. He will then compare their architecture and training algorithms. There is growing evidence that interleaved training of new and old data during non-REM sleep is essential, a strategy that works well with artificial neural networks as well. Take home messages: The hippocampus & cortex are analogous to a U-net, the cortex is similar to a conv-net, while the hippocampus is similar to a recurrent net, and our brains train during non-REM sleep. Could we improve generalization on fewer samples if we adopted biomimetic architectures and training algorithms?

Lightning Round Presentation: 8:05-8:15 Open

Networking - 8:15-9:00

All times are approximate. If you would like to present a demonstration or lightning round, please reach out to a coordinator to get it on the schedule.