Research map / Image generation / Fast diffusion sampling
On Unbiased Estimation for Partially Observed Diffusions
Jeremy Heng, Jeremie Houssineau, Ajay Jasra
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 the proposed maximal coupled resampling for the 4-CCPF instead of common uniform random variables (ablation) | unbiased score estimator · Ornstein-Uhlenbeck (T=25) | stopping time and increment variance | Improved |
| add burn-in b at the 90% quantile of stopping time and time-averaging window I=10b (ablation) | increment estimator · Ornstein-Uhlenbeck (T=25) | variance of increments | Improved |
| scale number of particles N linearly with number of observations T (ablation) | coupled conditional particle filter · Ornstein-Uhlenbeck (T growing) | stopping time | Improved |
| add adaptive resampling (when ESS < N/2) (ablation) | coupled conditional particle filter · red kangaroo logistic diffusion | dependence between CPF chains / stopping time | Improved |
| naive versus simple estimator of increments (ablation) | increment estimator · red kangaroo logistic diffusion | variance decay rate | No clear effect |
Closest papers by meaning
- Statistical Spatially Inhomogeneous Diffusion Inference (AAAI 2024 · similarity 0.88)
- Diffusion Path Samplers via Sequential Monte Carlo (NeurIPS 2026 · similarity 0.88)
- Learning Diffusion Priors from Observations by Expectation Maximization (NeurIPS 2024 · similarity 0.87)
- O(d/T) Convergence Theory for Diffusion Probabilistic Models under Minimal Assumptions (ICLR 2025 · similarity 0.87)
- Towards Non-Asymptotic Convergence for Diffusion-Based Generative Models (ICLR 2024 · similarity 0.87)
- Convergence of Score-Based Discrete Diffusion Models: A Discrete-Time Analysis (ICLR 2025 · 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.
