Reproduce “On Sparse Graph Limits for Neural Combinatorial Optimization” 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
Thien Le, Melanie Weber
The authors characterize what graph information supports transfer of normalized optimization values across sizes. Scalar energy probes preserve affine task values through the convex hull of assignment profiles. On hyperfinite graphs, attainable profiles approach that hull as graph size grows. However, constrained values can differ even when scalar features agree, as balanced-cut values do for K₄ and two disjoint copies of K₄.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.