Reproduce “Activation Functions, Statistics and Learning of Higher-Order Interactions in Restricted Boltzmann Machines” 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
Giovanni di Sarra, Yasser Roudi
The authors study how hidden-unit activation functions shape the visible-unit interactions produced by restricted Boltzmann machines (RBMs). They compare Linear, Step, ReLU, and Exponential activations using random parameter ensembles and training on a pure three-body target. Most random models favor lower-order interactions; near an Exponential regime, training better preserves the target structure than Step or ReLU in the tested setting.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.