Reproduce “Scaling of learning time for high dimensional inputs” 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

Carlos Stein Brito

The authors analyze learning by one neuron of a sparse feature in whitened high-dimensional data using nonlinear Hebbian learning. They track weight-feature overlap to estimate optimal-rate learning time. Theoretical scaling is N³/(log K)² for symmetric features and N²/log K for asymmetric features. Simulations support only the asymmetric scaling, with K=N and time measured until overlap reaches 0.7.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.