AI Paper Atlas

Research map / Optimization / Optimizers

Efficient Learning of Generative Models via Finite-Difference Score Matching

Tianyu Pang, Kun Xu, Chongxuan LI, Yang Song, Stefano Ermon, Jun Zhu

NeurIPS 2020 · Optimization › Optimizers · approach: Surrogate losses and robust solvers · 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
replace gradient-based SSM with finite-difference FD-SSMSSM · deep EBM, MNISTspeedup: 2.9xImproved
replace gradient-based DSM with finite-difference FD-DSMDSM · deep EBM, MNISTspeedup: 1.7xImproved
FD-DSM vs DSM final SM lossDSM · deep EBM, MNISTSM lossHurt
FD-SSM vs SSM on DKEF training timeSSM · DKEF Parkinsonstime per iteration: 110 ms -> 82.5 msImproved
FD-SSM vs SSM on DKEF test NLLSSM · DKEF ParkinsonsNLL: 14.52 -> 13.69Improved
FD-SSM vs SSM on DKEF test NLLSSM · DKEF RedWineNLL: 13.34 -> 13.06Improved
FD-SSM vs SSM on DKEF test NLLSSM · DKEF WhiteWineNLL: 14.13 -> 14.10No clear effect
FD-SSM vs SSMVR on DKEF test NLLSSMVR · DKEF ParkinsonsNLL: 13.26 -> 13.69Hurt
vary finite-difference step epsilon 0.1, 0.05, 0.02, 0.01 (ablation)FD-SSM · DKEF Parkinsonstest NLL: 14.17, 13.51, 14.03, 14.00No clear effect
FD-SSMVR vs SSMVR for VAE implicit encoder training timeSSMVR · VAE MNISTtime: 5.04 ms -> 3.98 msImproved
FD-SSMVR vs SSMVR for VAE implicit encoder training timeSSMVR · VAE CelebAtime: 14.9 ms -> 9.38 msImproved
FD-SSMVR vs SSMVR for VAE sample qualitySSMVR · VAE CelebAFID: 62.76 -> 64.85Hurt

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