Reproduce “From Boltzmann Machines to Neural Networks and Back Again” 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
Surbhi Goel, Adam Klivans, Frederic Koehler
The authors connect restricted Boltzmann machines (RBMs) and neural networks to learn structure and classify labels. A visible variable’s conditional mean is exactly a two-layer network with a tanh-like activation. Under bounded interaction norms and detectable two-hop neighbors, their method recovers visible-node Markov blankets; under separate nonnegative-interaction assumptions, an RBM classifier achieves 88.84 ± 0.31% FashionMNIST test accuracy.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.