Ferenc Huszár • 11/20/2020

Some Intuition on the Neural Tangent Kernel

This article provides an intuitive explanation of Neural Tangent Kernels (NTK), a tool for understanding neural network training dynamics. It uses a step-by-step 1D regression example, comparing a lookup table model to linear regression, to show how gradient descent updates affect the entire function, illustrating the core idea behind NTK and implicit regularization in wide neural networks.

0 comments

#Machine Learning #Gradient Descent #Neural Networks