Reproduce “Higher-Order Positional Encodings for Graph Representation Learning” 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
Caleb Stam, Aagrim Hoysal, Sanjukta Krishnagopal
The authors introduce preprocessing that computes Hodge positional features on a higher-order structure, projects them to nodes, and supplies them to an unchanged Graph Transformer. On a fixed-graph test with independently supplied filled triangles, Hodge predicts component counts with validation MAE 0.307 ± 0.013, versus 1.036 ± 0.0003 for LapPE. This shows transmission of supplied higher-order information, not discovery from graph data.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.