Reproduce “A method to automatically discover symbolic local learning rules” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Andrea Perin, Fabio Anselmi, Stephane Deny
The authors discover layer-specific local learning rules by learning coefficients over dictionaries of tensor expressions, then use each rule to update its layer’s weights. In small three-layer networks trained on Sliced MNIST, degree-2 rules outperform frozen-layer backpropagation on held-out digit classes, but not PCA or full backpropagation. The results do not establish scaling beyond these experiments.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.