Reproduce “Adaptive Annealed Importance Sampling with Constant Rate Progress” 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
Shirin Goshtasbpour, Victor Cohen, Fernando Perez-Cruz
CR-AIS adapts annealed importance sampling (AIS) schedules to estimate target normalization constants when densities are known only up to a constant. It uses weighted particles and divergence estimates to set each bridge step, avoiding ESS/CESS searches. In a 128-dimensional Student-t example, it reports 0.69 ± 0.27 log-normalizer error versus 1.37 ± 0.24, but is more accurate in only five of eight cases.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.