Research map / Probabilistic and scientific modeling / Variational and generative sampling
Kernel Stein Discrepancy thinning: a theoretical perspective of pathologies and a practical fix with regularization
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 entropic log-density regularization to KSD (regularized Stein thinning) | Stein thinning · Breast W. Bayesian logistic regression, NUTS, m=50 | AUC: 0.88 -> 0.96 | Helped |
| add entropic regularization to KSD | Stein thinning · Liver Bayesian logistic regression, NUTS, m=50 | AUC: 0.53 -> 0.69 | Helped |
| add entropic regularization to KSD | Stein thinning · Breast W. Bayesian logistic regression, MALA, m=50 | AUC: 0.68 -> 0.93 | Helped |
| add entropic regularization to KSD | Stein thinning · Liver Bayesian logistic regression, MALA, m=300 | AUC: 0.57 -> 0.62 | Helped |
| add entropic regularization to KSD | Stein thinning · Diabetes Bayesian logistic regression, NUTS, m=50 | AUC: 0.52 -> 0.50 | Hurt |
| add entropic regularization to KSD | Stein thinning · Diabetes Bayesian logistic regression, MALA, m=50 | AUC: 0.51 -> 0.48 | Hurt |
| add entropic regularization to KSD | Stein thinning · Sonar Bayesian logistic regression, NUTS, m=300 | AUC: 0.81 -> 0.81 | No clear effect |
| add entropic regularization to KSD | Stein thinning · Sonar Bayesian logistic regression, MALA, m=50 | AUC: 0.80 -> 0.80 | No clear effect |
| add entropic regularization to KSD | Stein thinning · Haberman Bayesian logistic regression, NUTS, m=300 | AUC: 0.53 -> 0.58 | Helped |
| add entropic regularization to KSD | Stein thinning · Diabetes Bayesian logistic regression, NUTS, m=300 | AUC: 0.53 -> 0.57 | Helped |
| use slower regularization rate lambda = 1/log(m) instead of 1/m (ablation) | regularized Stein thinning · 4-mode Gaussian mixture, MALA, MMD | MMD | Hurt |
| use faster regularization rate lambda = 1/m^2 instead of 1/m (ablation) | regularized Stein thinning · 4-mode Gaussian mixture, MALA, MMD | MMD | No clear effect |
Closest papers by meaning
- Generalized Kernel Thinning (ICLR 2022 · similarity 0.90)
- Measuring Sample Quality with Kernels (ICML 2017 · similarity 0.90)
- Debiased Distribution Compression (ICML 2024 · similarity 0.88)
- The Polynomial Stein Discrepancy for Assessing Moment Convergence (ICML 2025 · similarity 0.88)
- Supervised Kernel Thinning (NeurIPS 2024 · similarity 0.88)
- Quasi-Monte Carlo Features for Kernel Approximation (ICML 2024 · similarity 0.84)
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.