Reproduce “Scalable MCMC for Mixed Membership Stochastic Blockmodels” 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

Wenzhe Li, Sungjin Ahn, Max Welling

The authors extend stochastic-gradient Riemannian Langevin sampling to inference in assortative mixed-membership stochastic blockmodels, using mini-batches of node pairs and analytically marginalizing community assignments. For large community counts, SGMC-M groups low-probability states and samples their contribution. On RELATIVITY with K=50, both SGMC methods achieve much lower held-out perplexity than SVI; this comparison does not establish universal superiority.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.