AI Paper Atlas

Research map / Learning theory and algorithms / Optimal transport

HOT: An Efficient Halpern Accelerating Algorithm for Optimal Transport Problems

Guojun Zhang, Z. N. Gu, Yancheng Yuan, Defeng Sun

TPAMI 2025 · Learning theory and algorithms › Optimal transport · approach: Sinkhorn and entropic OT algorithms

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
use HOT instead of Network Simplexreduced OT solver · DOTmark Classic 256x256time (s): 2562.92 -> 12.98Improved
use HOT instead of Network Simplexreduced OT solver · DOTmark 128x128speedup: 17.44x speedupImproved
use HOT instead of Gurobi IPMreduced OT solver · DOTmark 128x128speedup: 15.83x speedupImproved
use HOT instead of convolutional SinkhornOT solver · DOTmark 128x128speedup: 19.54x speedupImproved
add Halpern acceleration (HOT vs generalized ADMM) (ablation)ADMM on reduced OT model · DOTmark 512x512iterations: saves 40% of iterations; 1610 -> 900 (Classic)Improved
add Halpern acceleration (HOT vs generalized ADMM) (ablation)ADMM on reduced OT model · DOTmark Classic 512x512time (s): 116.92 -> 81.02Improved
use HOT instead of ADMMreduced OT solver · DOTmark Shapes 64x64time (s): 3.92 -> 0.64Improved
use HOT instead of Network Simplex at small scalereduced OT solver · DOTmark Shapes 128x128gap: 2.46E-09 -> 2.51E-03Hurt
increase Sinkhorn regularization lambda from 0.01% to 1% of median cost (ablation)Convolutional Sinkhorn · DOTmark Classic 64x64time (s) and gap: time 16.18 -> 0.19; gap 1.69E-04 -> 3.29E-01Mixed
increase Sinkhorn regularization lambda from 0.01% to 0.1% of median cost (ablation)Convolutional Sinkhorn · DOTmark Classic 128x128gap: 6.98E-04 -> 3.34E-02Hurt
replace sparse Cholesky with the explicit linear-time linear-system solution (ablation)HOT subproblem · linear system solves over image sizessolve timeImproved

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