Research map / Image generation / Fast diffusion sampling
Denoising MCMC for Accelerating Diffusion-Based Generative Models
Beomsu Kim, Jong Chul Ye
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 changed | Where | Measured | Effect |
|---|---|---|---|
| Apply DLG initialization before KAR1 denoising | KAR1 score-based sampling · CIFAR10 | FID: 3.86 FID with 10.11 NFE; baseline 4.17 FID with 10 NFE | Improved |
| Apply DLG initialization before KAR1 denoising | KAR1 score-based sampling · CIFAR10 | FID: 2.63 FID with 20.11 NFE; baseline 2.86 FID with 20 NFE | Improved |
| Apply DLG initialization before KAR2 denoising | KAR2 score-based sampling · CelebA-HQ-256 | FID: 6.99 FID with 158.96 NFE; prior result 7.16 FID with 4000 NFE | Improved |
| Combine DLG with a reverse-S/ODE integrator | Six reverse-S/ODE integrators · CIFAR10 | FID | Improved |
| Combine DLG with KAR1 or RK45 | KAR1 and RK45 · CIFAR10 | FID: around 4 FID to around 2.4 FID | Improved |
| Run MCMC in data and noise-level product space using DLG | Langevin sampling on a 1k-mode Gaussian mixture · Mixture of Gaussians | mode coverage: all 1k modes at chain length 432 | Improved |
| Increase n_den/n from an extreme value toward an intermediate value (ablation) | DLG with deterministic integrator · CIFAR10 | FID | Improved |
| Increase n_den/n as NFE increases beyond 50 (ablation) | DLG with deterministic integrator · CIFAR10 | FID | Improved |
| Use an overly small or overly large Langevin step size (ablation) | DLG · CIFAR10 | FID | Hurt |
Closest papers by meaning
- Denoising Diffusion Bridge Models (ICLR 2024 · similarity 0.90)
- Particle Denoising Diffusion Sampler (ICML 2024 · similarity 0.90)
- Reverse Transition Kernel: A Flexible Framework to Accelerate Diffusion Inference (NeurIPS 2024 · similarity 0.90)
- O(d/T) Convergence Theory for Diffusion Probabilistic Models under Minimal Assumptions (ICLR 2025 · similarity 0.90)
- Accelerating Convergence of Score-Based Diffusion Models, Provably (ICML 2024 · similarity 0.90)
- Diffusion Bridge Implicit Models (ICLR 2025 · similarity 0.90)
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.
