Implicit Regularization in Deep Learning: The Kernel and Rich Regimes

Senior Project for Statistics & Data Science at Yale University completed by Henry Smith

I am grateful to have been advised by Dr. Harrison Zhou

Code Requirements: Python 3.7-3.9, TensorFlow 2.8

View the final project report and poster.

For proofs of the main results from "On Lazy Training in Differentiable Programming" by Chizat, Oyallon, and Bach, see another report I completed for S&DS 432, Advanced Optimization Techniques.