Reproduce “Quantifying Uncertainty In Wide Two-Layer Neural Networks: On The Law Of The Limiting Fluctuation Process” 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
Arnaud Descours, Arnaud Guillin, Geoffrey Lacour, Manon Michel, Boris Nectoux, Paul Stos
The authors derive an explicit covariance for prediction fluctuations in wide two-layer neural networks using a backward transport equation. The equation propagates an output observable backward along deterministic mean-field training dynamics and accounts for variation from initialization and training data. In a cubic-regression example, estimates qualitatively track a two-peaked Monte Carlo profile, without demonstrating computational savings or generality.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.