Research map / Probabilistic and scientific modeling / Variational and generative sampling
Stein Self-Repulsive Dynamics: Benefits From Past Samples
Mao Ye, Tongzheng Ren, Qiang Liu
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 |
|---|---|---|---|
| add Stein repulsive gradient to Langevin dynamics | Langevin dynamics · UCI Boston | test RMSE: 3.342 -> 3.086 | Improved |
| add Stein repulsive gradient to Langevin dynamics | Langevin dynamics · UCI Yacht | test log-likelihood: -1.187 -> -0.458 | Improved |
| add Stein repulsive gradient to Langevin dynamics | Langevin dynamics · UCI WineWhite | test RMSE: 0.692 -> 0.688 | No clear effect |
| add repulsive gradient to Langevin dynamics | Langevin dynamics · UCI Naval | test log-likelihood: 4.041 -> 4.186 | Improved |
| add repulsive gradient to Langevin dynamics | Langevin dynamics · Mushroom bandit | cumulative regret: 4.28 -> 3.80 | Improved |
| add repulsive gradient to Langevin dynamics | Langevin dynamics · Wheel bandit | cumulative regret: 38.07 -> 32.08 | Improved |
| use SRLD instead of SVGD | SVGD · Wheel bandit | cumulative regret: 91.32 -> 32.08 | Improved |
| use SRLD instead of SVGD | SVGD · UCI Concrete | test log-likelihood: -5.500 -> -3.034 | Improved |
| use SRLD instead of SVGD | SVGD · UCI WineWhite | test RMSE: 0.655 -> 0.688 | Hurt |
| increase SRLD samples used from 20 to 100 (ablation) | SRLD · UCI Boston | test RMSE: 3.236 -> 3.086 | Improved |
| increase SRLD samples used from 20 to 100 (ablation) | SRLD · UCI Yacht | test log-likelihood: -1.160 -> -0.458 | Improved |
| increase SRLD samples used from 20 to 100 (ablation) | SRLD · UCI WineWhite | test RMSE: 0.695 -> 0.688 | No clear effect |
Closest papers by meaning
- De-randomizing MCMC dynamics with the diffusion Stein operator (NeurIPS 2021 · similarity 0.89)
- A Dynamical System View of Langevin-Based Non-Convex Sampling (NeurIPS 2023 · similarity 0.89)
- High-Order Langevin Diffusion Yields an Accelerated MCMC Algorithm (JMLR 2021 · similarity 0.88)
- The Polynomial Stein Discrepancy for Assessing Moment Convergence (ICML 2025 · similarity 0.87)
- Fast Non-Log-Concave Sampling under Nonconvex Equality and Inequality Constraints with Landing (NeurIPS 2025 · similarity 0.86)
- The Poisson Midpoint Method for Langevin Dynamics: Provably Efficient Discretization for Diffusion Models (NeurIPS 2024 · similarity 0.86)
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.
