A learning algorithm for cortex

How does the brain learn its internal representations?

Causal generative models

Approximate inference

Energy-based models

The standard way to find the serious rivals

A very surprising shortcut

Slide 8

Slide 9

Autoencoders as generative models

Ways to restrict the code

A learning module

Bottom-up and top-down weights can be learned without considering the partition function

A picture of the learning procedure


Slide 16

Slide 17