Sample Complexity for the 2-Gromov-Wasserstein Distance

arXiv:2607.27514 2026 Geometry 1 ideas extracted · analyzed Aug 31, 2026

What the math gives to ML

The paper provides an explicit Hilbert-space encoding of Euclidean geometry: each point is mapped to its centered half-space incidence function, and squared Hilbert distance exactly equals the original Euclidean distance. This is more useful for neural networks than a generic GW estimator because it converts pairwise geometric comparison into ordinary feature-space operations and naturally supports objects living in different ambient dimensions. A practical transfer is a random half-space geometry layer for cross-domain alignment or graph learning, with empirical centering that removes distribution-dependent offsets and a controllable approximation error from the number of sampled half-spaces.

Ideas from this paper

Unverified 2026

Random Half-Space Geometry Layer

Represent each Euclidean input point by its responses to randomly sampled half-spaces, then center those responses by the minibatch or source-distribution half-space occupancy. Use squared distances between these representations as a geometry-preserving substitute for raw Euclidean distances in graph-NN edge construction, cross-domain retrieval, or geometry-aware attention. The layer can compare point clouds from different ambient dimensions because each domain has its own half-space dictionary…

Useful6/10
Difficulty4/10
Novelty6/10
Paper: Sample Complexity for the 2-Gromov-Wasserstein Distance arXiv:2607.27514