Linearizing Neural Networks
Greg and Daniel discuss the concept of linearizing neural networks using the neurotangent kernel, which simplifies the evolution of neural networks during training and allows for optimization and generalization results. They explain how this linearization relates to the ordinary kernel trick and the potential applications of this concept across different architectures.In this clip
From this podcast
The Gradient
Greg Yang on Communicating Research, Tensor Programs, and µTransfer
Related Questions