Research map / Probabilistic and scientific modeling / Variational and generative sampling
Using Perturbation to Improve Goodness-of-Fit Tests based on Kernelized Stein Discrepancy
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 MH-type jump perturbation (sum over grid of kernels, spKSD) to KSD | KSD test on bimodal Gaussian mixture with mixing-weight difference · 1D Gaussian mixture, mode separation large | rejection rate: from near 0.05 to almost perfect power | Helped |
| Use standard KSD on well-separated bimodal mixtures with differing mixing weights | KSD test · 1D Gaussian mixture, mode separation large | rejection rate: power converges to test level 0.05 | Hurt |
| Use optimised jump scale perturbation (ospKSD) instead of standard KSD on RAM samples with missing modes | RAM sample GOF test at sigma=0.5 | rejections out of 10: ospKSD 0 vs KSD 0 at sigma=0.5; ospKSD 8 vs KSD 0 at sigma=0.1; ospKSD 8 vs KSD 10 at sigma=1.08 | Mixed |
| Use grid-based spKSD instead of standard KSD on RAM samples | RAM sample GOF test · RAM scale sigma=0.1 | rejections out of 10: spKSD 0 vs KSD 0 | No clear effect |
Closest papers by meaning
- A Kernelized Stein Discrepancy for Goodness-of-fit Tests (ICML 2016 · similarity 0.92)
- KSD Aggregated Goodness-of-fit Test (NeurIPS 2022 · similarity 0.91)
- A Kernel Stein Test of Goodness of Fit for Sequential Models (ICML 2023 · similarity 0.91)
- The Polynomial Stein Discrepancy for Assessing Moment Convergence (ICML 2025 · similarity 0.89)
- Kernel Selection is Model Selection: A Unified Complexity-Penalised Approach for MMD Two-Sample Tests (NeurIPS 2026 · similarity 0.85)
- Sequential Kernel Goodness-of-fit Testing (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.