Reproduce “Convex Analysis of the Mean Field Langevin Dynamics” 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

Atsushi Nitanda, Denny Wu, Taiji Suzuki

The authors analyze convergence of mean-field neural-network training by comparing the current parameter distribution with a proximal Gibbs distribution built from the objective’s functional derivative. Under convexity and log-Sobolev assumptions, this comparison bounds the objective gap and yields geometric convergence in continuous time. For discrete noisy updates, they prove an accuracy-dependent rate with discretization error, not a finite-width guarantee.

Summary by alphaXiv.

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