Reproduce “Rènyi-Regularized Flow-Based Variational Inference for Bayesian Calibration” 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
Exaucè Luweh Adjim Ngarti, Élise Arnaud, Léo Nicoletti, Victor Trappler, Arthur Vidard
The authors add a prior-to-approximation Rényi penalty to the negative evidence lower bound for variational calibration with normalizing flows. In a two-parameter problem with separated posterior modes, the penalty encourages coverage of prior-supported regions. In the matched-flow comparison at ρ = 0.8 and r = 0.2, the near-collapse fraction falls from 0.50 to 0.25, but other metrics move in both directions and some runs still nearly collapse.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.