AI Paper Atlas

Research map / Image generation / GANs

Improved Techniques for Training Score-Based Generative Models

Yang Song, Stefano Ermon

NeurIPS 2020 · Image generation › GANs · approach: Stabilized GAN training · 138 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 NCSNv2 techniques instead of vanilla NCSNScore-based generative modeling · CelebA 64×64HYPE∞ (%): 19.8 -> 37.3Improved
Add EMA to the original NCSN (ablation)NCSN · CIFAR-10 and CelebAFID stabilityImproved
Add EMA and the noise-scale, conditioning, and sampler-configuration technique group (ablation)NCSN · CIFAR-10 and CelebAsample quality (visual inspection)Improved
Use all proposed techniques together rather than vanilla NCSN (ablation)NCSN · CIFAR-10 and CelebAFIDImproved
Use NCSNv2 rather than NCSNScore-based generative modeling · CIFAR-10Inception scoreHurt
Use NCSNv2 rather than NCSNScore-based generative modeling · CelebA 64×64HYPE∞ (%): 19.8 -> 37.3Improved

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