Research map / Learning theory and algorithms / Sample complexity
Improved Analysis of Score-based Generative Modeling: User-Friendly Bounds under Minimal Smoothness Assumptions
Hongrui Chen, Holden Lee, Jianfeng Lu
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 exponential integrator instead of Euler-Maruyama discretization | SGM sampler (trajectory-smooth setting) · KL(p0||q_T) bound | error bound terms | Improved |
| Exponentially decreasing step size instead of uniform discretization (ablation) | SGM with early stopping · discretization term Pi, dependence on 1/delta | scaling of discretization error | Improved |
| Linear step size (quadratic discretization points) instead of exponentially decreasing step size (ablation) | SGM with early stopping · discretization term Pi, dependence on 1/delta | scaling of discretization error | Mixed |
| Replace uniform Hessian bound with high-probability Hessian bound and change of measure | SGM for general distributions with early stopping · steps for TV error vs prior bounded-support result | number of steps | Improved |
| Add truncation step to sampler output | SGM pure Wasserstein guarantee · W2 guarantee depending on tail decay | Wasserstein distance | Improved |
| Use reverse KL guarantee instead of TV guarantee for trajectory-smooth setting | SGM analysis · trajectory-smooth | guarantee strength | Improved |
Closest papers by meaning
- Beyond Log-Concavity and Score Regularity: Improved Convergence Bounds for Score-Based Generative Models in W2-distance (ICML 2025 · similarity 0.92)
- Global Well-posedness and Convergence Analysis of Score-based Generative Models via Sharp Lipschitz Estimates (ICLR 2025 · similarity 0.92)
- Score-based generative models are provably robust: an uncertainty quantification perspective (NeurIPS 2024 · similarity 0.91)
- Algorithm- and Data-Dependent Generalization Bounds for Diffusion Models (NeurIPS 2025 · similarity 0.89)
- Sampling is as easy as keeping the consistency: convergence guarantee for Consistency Models (ICML 2024 · similarity 0.89)
- Approximation and Generalization Abilities of Score-based Neural Network Generative Models for Sub-Gaussian Distributions (NeurIPS 2025 · 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.
