AI Paper Atlas

Research map / Point clouds / Point cloud classification

Globally solving the Gromov-Wasserstein problem for point clouds in low dimensional Euclidean spaces

Martin Ryner, Jan Kronqvist, Johan Karlsson

NeurIPS 2023 · Point clouds › Point cloud classification · approach: Equivariant point cloud networks · 3 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
solve concave subproblem by extreme point search instead of Gurobi spatial branch and bound (ablation)Algorithm 1 · U n=10 dims 2,2time (s): 21 (6-47) -> 0.14 (0.07-0.3)Improved
extreme point search instead of B&B (ablation)Algorithm 1 · U n=500 dims 2,2time (s): 408 (269-511) -> 11 (9-16)Improved
extreme point search instead of B&B (ablation)Algorithm 1 · N1 n=10 dims 2,3time (s): 708 (233-1184) -> 0.51 (0.39-0.65)Improved
extreme point search instead of B&B (ablation)Algorithm 1 · N3 n=2000 dims 3,3time (s): 578 (429-691) -> 93 (91-100)Improved
proposed method vs MILP1 formulationAlgorithm 1 · U n=10 dims 2,2time (s): 39 (11-58) -> 0.14 (0.07-0.3)Improved
proposed method vs direct Gurobi on larger problemsAlgorithm 1 · U n=100 dims 2,2time (s): 25 (19-39) -> 0.48 (0.3-0.7)Improved
proposed method vs direct Gurobi on tiny 3D problemsAlgorithm 1 · N2 n=10 dims 3,3time (s): 1.8 (0.7-3.2) vs 0.2(0.1-0.3)Hurt
proposed vs direct Gurobi at n>=500Algorithm 1 · U n=500 dims 2,2time (s): - -> 11 (9-16)Improved
increase point dimension from 2,2 to 2,3Algorithm 1 · U n=100time (s): 0.48 (0.3-0.7) -> 278 (99-813)Hurt
reflection-dominated data (N3) vs rotation-rich data (N2)Algorithm 1 · n=500 dims 3,3time (s): 436 (228-862) -> 11 (9-16)Improved
proposed method vs local search with random restartsAlgorithm 1 · U n=400 dims 2,2time (s): 322 (97-685) -> 6 (5-8)Improved
proposed method vs local searchAlgorithm 1 · N1 n=400 dims 2,2time (s): 149 (22-378) -> 4.4 (3.6-5.3)Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy