Research map / Image generation / Fast diffusion sampling
Metropolis Sampling for Constrained Diffusion Models
Nic Fishman, Leo Klarner, Emile Mathieu, Michael John Hutchinson, Valentin De Bortoli
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 changed | Where | Measured | Effect |
|---|---|---|---|
| use Metropolis forward process instead of reflected Brownian motion | constrained diffusion model · hypercube d=2 | test log-likelihood: 2.25 -> 2.32 | Improved |
| use Metropolis instead of reflected forward process | constrained diffusion model · hypercube d=3 | test log-likelihood: 3.77 -> 4.15 | Improved |
| use Metropolis instead of reflected forward process | constrained diffusion model · hypercube d=10 | test log-likelihood: 7.42 -> 10.80 | Improved |
| use Metropolis instead of reflected forward process | constrained diffusion model · simplex d=3 | test log-likelihood: 2.64 -> 3.23 | Improved |
| use Metropolis instead of reflected forward process | constrained diffusion model · simplex d=10 | test log-likelihood: 7.00 -> 7.81 | Improved |
| use Metropolis instead of reflected forward process | constrained diffusion model · hypercube d=10 | training time (hours): 10.1 -> 0.90 | Improved |
| increase dimensionality from 2 to 10 | Metropolis vs reflected gap · hypercube | test log-likelihood: 2.25 -> 2.32 at d=2; 7.42 -> 10.80 at d=10 | Improved |
| use Metropolis instead of reflected forward process | constrained diffusion model · Robotics (SPD x R^2) | test log-likelihood: 8.39 -> 9.13 | Improved |
| use Metropolis instead of reflected forward process | constrained diffusion model · Robotics (SPD x R^2) | training time (hours): 9.52 -> 1.36 | Improved |
| use Metropolis instead of reflected forward process | constrained diffusion model · Protein backbones | test log-likelihood: 15.20 -> 15.33 | Improved |
| use Metropolis instead of reflected forward process | constrained diffusion model · Protein backbones | training time (hours): 24.80 -> 3.12 | Improved |
| add boundary constraint via Metropolis instead of unconstrained diffusion | Metropolis diffusion · Wildfire data on S^2 | MMD: 0.1567 -> 0.1388 | Improved |
Closest papers by meaning
- Fast constrained sampling in pre-trained diffusion models (NeurIPS 2025 · similarity 0.90)
- Diffusion Bridge Implicit Models (ICLR 2025 · similarity 0.89)
- CCS: Controllable and Constrained Sampling with Diffusion Models via Initial Noise Perturbation (NeurIPS 2025 · similarity 0.89)
- Constrained Diffusion with Trust Sampling (NeurIPS 2024 · similarity 0.89)
- Adapting to Unknown Low-Dimensional Structures in Score-Based Diffusion Models (NeurIPS 2024 · similarity 0.89)
- A Variational Perspective on Solving Inverse Problems with Diffusion Models (ICLR 2024 · similarity 0.89)
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.
