AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Denoising MCMC for Accelerating Diffusion-Based Generative Models

Beomsu Kim, Jong Chul Ye

ICML 2023 · Image generation › Fast diffusion sampling · approach: Denoising diffusion probabilistic models · 4 citations (OpenAlex)

Read the paper

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
Apply DLG initialization before KAR1 denoisingKAR1 score-based sampling · CIFAR10FID: 3.86 FID with 10.11 NFE; baseline 4.17 FID with 10 NFEImproved
Apply DLG initialization before KAR1 denoisingKAR1 score-based sampling · CIFAR10FID: 2.63 FID with 20.11 NFE; baseline 2.86 FID with 20 NFEImproved
Apply DLG initialization before KAR2 denoisingKAR2 score-based sampling · CelebA-HQ-256FID: 6.99 FID with 158.96 NFE; prior result 7.16 FID with 4000 NFEImproved
Combine DLG with a reverse-S/ODE integratorSix reverse-S/ODE integrators · CIFAR10FIDImproved
Combine DLG with KAR1 or RK45KAR1 and RK45 · CIFAR10FID: around 4 FID to around 2.4 FIDImproved
Run MCMC in data and noise-level product space using DLGLangevin sampling on a 1k-mode Gaussian mixture · Mixture of Gaussiansmode coverage: all 1k modes at chain length 432Improved
Increase n_den/n from an extreme value toward an intermediate value (ablation)DLG with deterministic integrator · CIFAR10FIDImproved
Increase n_den/n as NFE increases beyond 50 (ablation)DLG with deterministic integrator · CIFAR10FIDImproved
Use an overly small or overly large Langevin step size (ablation)DLG · CIFAR10FIDHurt

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