Brains@Bay Alternatives to Backpropagation - Nov 18 9:30 am Pacific Time

Hi all!

As you know we started Brains@Bay last year as an attempt to get the machine learning community closer to the neuroscience world. It is our ultimate belief that the best way to develop machine intelligence is to understand our only proof of existence, the brain, and we want to bring as many people on board as possible.

At Brains@Bay we discuss a wide range of topics, some which are closely related to our theories at Numenta, and some less related. We like to keep an open mind and study each research topic from different points of view.

This Wednesday 9:30 a.m. Pacific Time we will host our last 2020 edition. This time we will talk about alternatives to backpropagation in neural networks. We have a great lineup of speakers, with Rafal Bogacz, Sindy Löwe and Jack Kendall. We expect great talks and a very interesting discussion in the panel.

Thank you all again for your support. Special thanks to @sheiser1 who was there from the beginning helping us organize our first editions. Hope to see you there!

5 Likes

Great meeting. Very interesting presentations.

Thanks to @lucasosouza for organizing this. Thanks to Numenta for sponsoring this. And many thanks to the speakers/researchers.

I found links to the papers the three speakers presented. (Please correct me if I’m wrong).

Prof. Rafal Bogacz, from University of Oxford
Can the Brain Do Backpropagation? Exact Implementation of Backpropagation in Predictive Coding Networks

(Here is a working link to the pdf)

Sindy Löwe from University of Amsterdam
Putting an End to End-to-End: Gradient-Isolated Learning of Representations

Jack Kendall, co-founder of RAIN Neuromorphics
Training End-to-End Analog Neural Networks with Equilibrium Propagation

Oh, and I almost forgot. Here is the rerun on YouTube:

4 Likes

Thanks @Falco. I would also add this review paper from James Whittington and Prof. Rafal Bogacz:
https://www.cell.com/trends/cognitive-sciences/fulltext/S1364-6613(19)30012-9

3 Likes