Reproduce “Variational Learning with Lie Group Exponential Families” 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
Eren Mehmet Kıral, Thomas Möllenhoff, Koichi Tojo, Mohammad Emtiyaz Khan, Kenichi Bannai
The authors introduce LG-BLR, a variational update that represents candidate distributions as transformations of a base distribution by a Lie group. It updates by composing the current group element with an exponential-map step, keeping the distribution within the chosen family. For open exponential families with transitive group actions, its first-order limit is BLR; curved families instead yield a Fisher-metric projection.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.