Reproduce “Gromov-Wasserstein Graph Coarsening” 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
Carlos A. Taveras, Santiago Segarra, César A. Uribe
Researchers at Rice University developed GPC and KGPC, novel graph coarsening algorithms that leverage Gromov-Wasserstein (GW) geometry to minimize structural distortion during graph size reduction. These methods demonstrated superior preservation of graph structure and achieved competitive, often better, performance in graph classification tasks compared to existing baselines and uncoarsened graphs.
Key insights
- Minimizing Gromov-Wasserstein distortion offers a robust, geometrically-aware framework for graph coarsening that can compare unaligned graphs of different sizes.
- GPC is theoretically guaranteed to recover the smallest network weakly isomorphic to the original, implying a form of “lossless” structural compression in the GW sense.
- GW-aware coarsening can not only reduce computational load but also potentially enhance downstream task performance by effectively preserving essential graph structure and possibly filtering noise.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.