This repo contains code for replicating the experiments of the paper The Eigenlearning Framework: A Conservation Law Perspective on Kernel Regression and Wide Neural Networks.

Selected experiments are contained in the experiments directory, each containing brief explanations and code to generate the figures seen in our paper. These experiments illustrate our theory and we recommend starting with these notebooks. They can be run in Google Colab or locally.

The .py files provide a general codebase for generating synthetic datasets, loading image datasets, and testing the performance of kernel regression and finite nets learning functions on these domains.

Please let us know if you run into any bugs or issues.