Salon with Maithra Raghu

Maithra presents on the differences between wide and deep networks. Made by Angelica Pan using Weights & Biases
Angelica Pan
This salon features Maithra Raghu of Google Brain, who presents work with Thao Nguyen and Simon Kornblith answering the question: "Do Wide and Deep Networks Learn the Same Things?".
She presents evidence that generic high-capacity networks learn different representations than low-capacity networks and that there are distinct differences between the predictions and the errors of wide vs deep models with the same accuracy.
The paper: