Reproduce “Generator-based Graph Generation via Heat Diffusion” 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
Anthony Stephenson, Ian Gallagher, Christopher Nemeth
G³ generates graph samples by learning to reverse a continuous heat-diffusion process guided by each training graph’s Laplacian. A neural network approximates the known generator, and Euler integration turns a symmetric base sample into a thresholded adjacency matrix. On Enzymes, its clustering-coefficient MMD is 0.0225 ± 0.0008, but results vary across datasets and properties, with no overall winner.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.