AI Paper Atlas

Research map / Optimization / Optimizers

Estimating the Rate-Distortion Function by Wasserstein Gradient Descent

Yibo Yang, Stephan Eckstein, Marcel Nutz, Stephan Mandt

NeurIPS 2023 · Optimization › Optimizers · approach: Stein variational gradient descent · 2 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
Use L_BA rather than L_EOT for WGD (ablation)WGD · R-D estimationcomputational speed: 10 to 100 times fasterImproved
Alternate WGD updates with BA weight updates (ablation)WGD · deconvolutionconvergence speedImproved
Use WGD instead of neural upper-bound methodsR-D estimation · MNISTconvergence speed: at least an order of magnitude fasterImproved
Use WGD instead of NERD at the same computation budgetR-D estimation · physics, speech, and MNISTR-D upper-bound tightnessImproved
Increase the number of particles (ablation)WGD · R-D estimationR-D upper-bound tightnessImproved
Move reproduction particles rather than keeping their initial locations fixed (ablation)particle-based R-D estimation · deconvolutionoptimality gapImproved
Use a particle-based method rather than the neural methodsdeconvolution · uniform measure on the unit circlereproduction-support accuracyImproved
Use WGD or its hybrid variant rather than BA or NERDR-D estimation · deconvolution across dimensions and lambda valuesoptimality gapImproved

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