Salon with Diganta Misra

Diganta talks about smooth activations, robustness, and catastrophic forgetting. Made by Angelica Pan using Weights & Biases
Angelica Pan
This salon features Diganta Misra, of Hong Kong University and Landskape, who presents some thoughts on the role of smooth non-linearities, like Swish and Mish, in the robustness of neural networks and in catastrophic forgetting.