Skip to contentSkip to filters

We use cookies for analytics.

Grokking, Generalization Collapse, and the Dynamics of Training Deep Neural Networks

Related content

About

Charles Martin discusses delayed generalization in deep neural networks. The conversation examines how the behavior of the weights during training relates to a model’s ability to generalize and why strong training performance alone can conceal a later collapse.

Topics

Details

Show

The TWIML AI Podcast

48 episodes on World Wide · TWIML