Reproduce “Contrastive Laplacian Eigenmaps” 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
Hao Zhu, Ke Sun, Piotr Koniusz
COLES extends Laplacian Eigenmaps by adding repulsion from sampled negative graphs alongside attraction between connected nodes. Its objective rewards positive dot products and penalizes negative ones; the authors relate it to a Wasserstein-distance surrogate under a Lipschitz condition. On transductive Cora classification with five labeled examples per class, COLES-GCN reaches 73.8 ± 3.4% accuracy versus 61.3 ± 4.3% for GCN with SampledNCE, limited to this setting.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.