Gradient Kernel Regression

Matt Calder

Gradient Kernel Regression

Artificial Intelligence 2021-04-14 v1

Authors: Matt Calder

Abstract

In this article a surprising result is demonstrated using the neural tangent kernel. This kernel is defined as the inner product of the vector of the gradient of an underlying model evaluated at training points. This kernel is used to perform kernel regression. The surprising thing is that the accuracy of that regression is independent of the accuracy of the underlying network.

Keywords

kernel learning neural networks neural network

Cite

@article{arxiv.2104.05874,
  title  = {Gradient Kernel Regression},
  author = {Matt Calder},
  journal= {arXiv preprint arXiv:2104.05874},
  year   = {2021}
}

Related papers

View all related →