Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

PINS: Proximal Iterations with Sparse Newton and Sinkhorn for Optimal Transport

About

Optimal transport (OT) is a widely used tool in machine learning, but computing high-accuracy solutions for large instances remains costly. Entropic regularization and the Sinkhorn algorithm improve scalability; however, when the regularization parameter is small, Sinkhorn convergence slows, and the iterates approach an entropic solution that remains separated from the true OT plan by an entropic-bias plateau. We introduce PINS (Proximal Iterations with sparse Newton and Sinkhorn), a two-loop solver designed to move beyond this plateau. The outer loop applies an entropic proximal-point method, solving the original OT problem through a sequence of entropic subproblems with shifted cost matrices. Each inner subproblem is then solved by a Sinkhorn warm-up followed by sparse-Newton refinement. We prove that PINS converges globally to an optimal solution of the unregularized OT problem and that the inner Hessian admits a sparsification at every outer iteration with a structure independent of the cost matrix. On synthetic and augmented-MNIST instances, PINS achieves much lower relative cost errors than Sinkhorn-type baselines, which stall at the entropic-bias plateau, and is $5$--$73\times$ faster than Sinkhorn with the same outer loop at matched accuracy. On large-scale DOTmark instances, a streaming implementation reduces peak memory by $24$--$54\%$ compared with the network-simplex linear programming (LP) solver and remains feasible under per-process memory budgets for which the LP solver fails.

Di Wu, Ling Liang, Haizhao Yang• 2025

Related benchmarks

TaskDatasetResultRank
Optimal TransportDOTmark
Execution Time (s)1
18
Optimal TransportSynthetic
Log10 Relative Error-4.75
12
Optimal TransportMNIST
Log10 Relative Error-3.48
12
Color transferAutumn–Grafiti
SSIM53
2
Color transferAutumn–Comunion
SSIM0.54
2
Color transferGrafiti–Comunion
SSIM76
2
Optimal TransportDOTmark 64
Peak RSS (GB)0.63
2
Optimal TransportDOTmark 128
Peak RSS (GB)5.29
2
Optimal TransportRandom Problems 15000
Peak RSS (GB)6.94
2
Optimal TransportRandom Problems random_20000
Peak RSS (GB)12.39
2
Showing 10 of 10 rows

Other info

Follow for update