arxivcs.LGq-bio.QMstat.ML2026-07-07
EntroPath: Maximum Entropy Path Ensemble Embedding for Manifold Learning
We introduce EntroPath, a manifold learning method that recovers geodesic geometry from data graphs through ensembles of diffusion paths. Many existing graph-based embeddings rely either on locally normalised random walks or on shortest-path distances. The former can concentrate…