Reproduce “Variational Inference of overparameterized Bayesian Neural Networks: a theoretical and empirical study” 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
Tom Huix, Szymon Majewski, Alain Durmus, Eric Moulines, Anna Korba
The authors analyze how width-dependent scaling affects variational inference in Bayesian neural networks. Their method sets the evidence lower bound’s KL penalty according to observation count and inversely to neuron count. Under assumptions, the unscaled objective approaches the prior as width grows. On CIFAR-10, ResNet20 accuracy falls at large trade-off values; this alone does not validate uncertainty estimates.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.