Research map / Image generation / Fast diffusion sampling
Diffusion Differentiable Resampling
Jennifer R. Andersson, Zheng Zhao
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 |
|---|---|---|---|
| increase diffusion discretisation from K=8 to K=128 and T from 1 to 3 (ablation) | diffusion resampling · Gaussian mixture resampling | sliced 1-Wasserstein distance (x10^-1): 1.64 -> 0.80 | Improved |
| use diffusion resampling instead of multinomial resampling | particle filter · Gaussian mixture resampling | resampling variance of mean estimator (x10^-2): 3.78 -> 3.74 | No clear effect |
| use diffusion resampling instead of multinomial resampling | particle filter · Linear Gaussian SSM | filtering KL (x10^-1): 5.49 -> 4.26 | Improved |
| use diffusion resampling instead of multinomial resampling | L-BFGS parameter estimation · Linear Gaussian SSM | parameter error (x10^-1): NaN -> 1.28 | Improved |
| use diffusion resampling instead of OT resampling | particle filter · Linear Gaussian SSM | log-likelihood function error: 2.64 -> 2.55 | Improved |
| use diffusion resampling instead of Gumbel-Softmax or Soft resampling | neural SSM learning · Prey-predator model | prediction RMSE | Improved |
| use diffusion resampling instead of OT, Gumbel, Soft | pendulum dynamics learning · Vision-based pendulum | SSIM and PSNR | No clear effect |
| use diffusion resampling instead of OT | resampling runtime · N=8192, K vs 1/epsilon | computation time | Improved |
Closest papers by meaning
- Correcting Diffusion Generation through Resampling (CVPR 2024 · similarity 0.89)
- Diffusing Differentiable Representations (NeurIPS 2024 · similarity 0.89)
- Particle Denoising Diffusion Sampler (ICML 2024 · similarity 0.89)
- A Variational Perspective on Solving Inverse Problems with Diffusion Models (ICLR 2024 · similarity 0.88)
- Solving Inverse Problems with Latent Diffusion Models via Hard Data Consistency (ICLR 2024 · similarity 0.87)
- Resfusion: Denoising Diffusion Probabilistic Models for Image Restoration Based on Prior Residual Noise (NeurIPS 2024 · similarity 0.87)
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.
