Research map / Image generation / Fast diffusion sampling
Infinite-Dimensional Generative Diffusions via Doob’s h-Transform
Thorben Pieper-Sethmacher, Daniel Paulin
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 |
|---|---|---|---|
| shorten noising time T from 1 to 0.2 for time-reversal HND (ablation) | HND · Gaussian mixture (D=500) | sliced Wasserstein distance: 0.117 -> 0.51 | Hurt |
| shorten noising time T from 1 to 0.2 for forced model FNS (ablation) | FNS · Gaussian mixture (D=500) | sliced Wasserstein distance: 0.055 -> 0.053 | No clear effect |
| shorten noising time T from 1 to 0.2 for ND (ablation) | ND · Gaussian mixture (D=500) | sliced Wasserstein distance: 0.274 -> 0.692 | Hurt |
| use linear noise schedule instead of constant noise in forced diffusion (ablation) | forced VP-SDE · Gaussian mixture (D=500, T=1) | sliced Wasserstein distance: 0.258 -> 0.055 | Improved |
| increase state dimension D with finite-dimensional ND model (ablation) | ND · Gaussian mixture | score approximation error | Hurt |
| use infinite-dimensional-aware models (forced, HND) across dimensions (ablation) | forced diffusion / HND · Gaussian mixture | score approximation error | No clear effect |
| smaller roughness parameter gamma near 0 (ablation) | VP-SPDE · Gaussian mixture and MNIST-SDF | performance | Improved |
| VP-SPDE vs DDO on MNIST-SDF | VP-SPDE · MNIST-SDF 64x64 | FID: 2.74 -> 3.12 | Hurt |
| forced VP-SPDE instead of noising-denoising baseline | noising-denoising · Seismic imaging | relative error: 0.17 -> 0.09 | Improved |
| forced VP-SPDE posterior uncertainty vs baseline | noising-denoising · Seismic imaging | average posterior std: 24.27 -> 35.28 | Mixed |
Closest papers by meaning
- Towards Non-Asymptotic Convergence for Diffusion-Based Generative Models (ICLR 2024 · similarity 0.90)
- DEFT: Efficient Fine-tuning of Diffusion Models by Learning the Generalised $h$-transform (NeurIPS 2024 · similarity 0.88)
- Bridging discrete and continuous state spaces: Exploring the Ehrenfest process in time-continuous diffusion models (ICML 2024 · similarity 0.88)
- A solvable model of learning generative diffusion: theory and insights (NeurIPS 2025 · similarity 0.88)
- Dimension-free Score Matching and Time Bootstrapping for Diffusion Models (NeurIPS 2025 · similarity 0.87)
- Lipschitz Singularities in Diffusion Models (ICLR 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.
