Research map / Learning theory and algorithms / Sample complexity
On the Robustness of Kernel Goodness-of-Fit Tests
Xing Liu, François-Xavier Briol
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 |
|---|---|---|---|
| increase tilt exponent b of the weighting function (ablation) | tilted-kernel KSD test · Toy Gaussian, outlier z=10 | rejection probability | Hurt |
| use a tilted kernel instead of a stationary IMQ kernel (ablation) | KSD test · Toy Gaussian with outliers | rejection probability under contamination | Improved |
| use the robust-KSD test instead of standard tilted KSD | tilted KSD · Toy Gaussian with contamination | Type-I error | Improved |
| aggregate over multiple bandwidths (KSDAgg) instead of robust-KSD (ablation) | robust KSD · Toy Gaussian with contamination | robustness | Hurt |
| choose bandwidth by median heuristic rather than fixed values (ablation) | robust-KSD · Toy Gaussian | power when epsilon exceeds epsilon_0 | Improved |
| use robust-KSD with radius for nu_0 = 5 on t-distributed data (ablation) | robust-KSD · misspecified tails | power | Hurt |
| use robust-KSD with theta for nu_0 = 10 or 20 (ablation) | robust-KSD · misspecified tails | power at small nu | Improved |
| use robust-KSD test on Gaussian-Bernoulli RBM with epsilon_0 = 0.1 | robust-KSD · Gaussian-Bernoulli RBM | Type-I error | Improved |
| use robust test instead of standard on kernel exponential family model with outliers | KSD tests · KEF galaxy data | Type-I error | Improved |
| increase mode separation gamma in Gaussian mixtures (ablation) | standard and robust KSD tests · mixture of Gaussians | power | Hurt |
Closest papers by meaning
- A Linear-Time Kernel Goodness-of-Fit Test (NeurIPS 2017 · similarity 0.91)
- A Kernel Test of Goodness of Fit (ICML 2016 · similarity 0.91)
- Using Perturbation to Improve Goodness-of-Fit Tests based on Kernelized Stein Discrepancy (ICML 2023 · similarity 0.90)
- Kernel Selection is Model Selection: A Unified Complexity-Penalised Approach for MMD Two-Sample Tests (NeurIPS 2026 · similarity 0.87)
- Sequential Kernel Goodness-of-fit Testing (ICML 2024 · similarity 0.86)
- Kernel conditional tests from learning-theoretic bounds (NeurIPS 2025 · 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.
