Reproduce “AgraSSt: Approximate Graph Stein Statistics for Interpretable Assessment of Implicit Graph Generators” 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
Wenkai Xu, Gesine Reinert
AgraSSt tests whether an implicit graph generator matches an observed target using samples from the generator. It estimates conditional edge probabilities relative to a chosen graph statistic, then compares the target with simulated graphs. Across 100 E2ST trials, its rejection rate for GraphRNN was 0.42, versus 0.04 for Monte Carlo; this supports the diagnostic in that setup, not universal rankings.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.