Spar-Sink applies importance sampling based on upper bounds of transport plans to create a sparse approximation of the Sinkhorn kernel, yielding consistent estimators for regularized OT and UOT with near-linear iteration cost.
Hilbert curve projection distance for distribution comparison
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
stat.ML 1years
2023 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Importance Sparsification for Sinkhorn Algorithm
Spar-Sink applies importance sampling based on upper bounds of transport plans to create a sparse approximation of the Sinkhorn kernel, yielding consistent estimators for regularized OT and UOT with near-linear iteration cost.