AI Paper Atlas

Research map / Optimization / Nonconvex and smooth optimization

Semi-Dual Unbalanced Quadratic Optimal Transport: fast statistical rates and convergent algorithm.

Adrien Vacher, François-Xavier Vialard

ICML 2023 · Optimization › Nonconvex and smooth optimization · approach: Langevin sampling for nonconvex problems · 1 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
Semi-dual formulation (strongly convex potentials) instead of Sinkhorn for unbalanced OT potential estimationUnbalanced OT potential estimation · 6D synthetic experiment, n=400 samplessquared error ||z_hat - z0||^2Improved
Increase strong convexity parameter lambda to 2.0 (beyond true 1-strong convexity of z0) (ablation)Semi-dual model · 6D synthetic experimentsquared error ||z_hat - z0||^2Hurt
Semi-dual model used in place of Sinkhorn for balanced 2D shape matchingSemi-dual OT (lambda=0.2, L=1) · 2D shape matchingqualitative sharpness of pushforwardImproved
Variable metric descent (semi-dual, Bregman-based) vs Frank-Wolfe in balanced 2D experimentSemi-dual optimization algorithm · Balanced potential recovery (Fig. 2)convergence speed (iterations)Improved
Exponential convergence for smooth strongly convex potentials (balanced) vs only strongly convex potentials (ablation)Variable metric descent on semi-dual · Balanced case, Fig. 3convergence of log squared L2 gradient errorImproved
Unbalanced case instead of balanced case for variable metric descent (ablation)Semi-dual optimization algorithm · Simple 1D example, Fig. 4convergence speedHurt
Semi-dual with convexity constraint vs SSNB non-convex objective in balanced 2D shape matchingSemi-dual OT vs SSNB · 2D shape matchingW2^2 between pushforward and targetImproved

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