AI Paper Atlas

Research map / Learning theory and algorithms / Optimal transport

Wasserstein Iterative Networks for Barycenter Estimation

Alexander Korotin, Vage Egiazarian, Lingxiao Li, Evgeny Burnaev

NeurIPS 2022 · Learning theory and algorithms › Optimal transport · approach: Wasserstein barycenters and GW · 4 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
replace ICNN-based SCW2B with WIN (generator output)barycenter estimation · Ave, celeba!FID: 156.3 -> 75.8Improved
replace SCW2B with WIN, weighted mapped outputbarycenter estimation · Ave, celeba!FID: 152.1 -> 52.85Improved
replace SCW2B with WIN for inverse maps, n=3inverse OT maps · Ave, celeba!FID: 319.62 -> 61.5Improved
replace SCW2B with WIN for inverse maps, n=1inverse OT maps · Ave, celeba!FID: 67.4 -> 49.3Improved
replace constant-shift baseline with WIN inverse maps, n=2inverse OT maps · Ave, celeba!FID: 75.5 -> 46.9Improved
replace SCW2B with WINbarycenter, Gaussian location-scatter D=128 · BW2-UVPBW2-UVP (%): 1.28 -> 0.38Improved
replace SCW2B with WINbarycenter, Uniform location-scatter D=128 · BW2-UVPBW2-UVP (%): 1.38 -> 0.46Improved
replace SCW2B with WINbarycenter, Gaussian location-scatter D=2 · BW2-UVPBW2-UVP (%): 0.07 -> 0.01Improved
replace ICNN-based SCW2B with WIN as N=1 generative modelCelebA generation · CelebA 64x64FID: 90.2 -> 46.6Improved
compare WIN to WGAN-QC as N=1 generative model (mapped output)CelebA generation · CelebA 64x64FID: 14.4 vs 15.7No clear effect
fix randomly initialized generator, train only the map (ablation)WIN, N=1 · CelebA 64x64FID: 15.7 -> 16.1No clear effect
map generated samples through learned input maps (weighted sum)WIN · Ave, celeba!FID: 75.8 -> 52.85Improved

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