Research map / Learning theory and algorithms / Prediction and decision losses
On the Exploration of Local Significant Differences For Two-Sample Test
Zhijian Zhou, Jie Ni, Jia-He Yao, Wei Gao
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 |
|---|---|---|---|
| MEMaBiD instead of ME (isotropic single Gaussian kernel) | MEMaBiD · Two-sample test, 16 datasets avg | test power: .795 -> .909 | Improved |
| MEMaBiD instead of MMDAgg | MEMaBiD · Two-sample test, 16 datasets avg | test power: .859 -> .909 | Improved |
| MEMaBiD instead of deep-kernel MMD-D | MEMaBiD · Two-sample test, 16 datasets avg | test power: .843 -> .909 | Improved |
| MEMaBiD instead of AutoMLTST | MEMaBiD · Two-sample test, 16 datasets avg | test power: .863 -> .909 | Improved |
| MEMaBiD instead of AutoMLTST on labor | MEMaBiD · labor | test power: 1.00±.000 vs .992±.012 | Hurt |
| MEMaBiD instead of AutoMLTST on poker | MEMaBiD · poker | test power: .832±.048 vs .821±.079 | Hurt |
| Proposed partition-based exploration instead of FDG | Our method · Local difference exploration, avg of 8 datasets | density difference: .907 -> .959 | Improved |
| Proposed exploration instead of MMDT on sea50 | Our method · sea50 | density difference: .977±.056 vs .985±.045 | Hurt |
| Proposed exploration instead of K-PRIM and BTLDD | Our method · Local difference exploration, avg | density difference: .891 -> .959 | Improved |
| Bi-directional hypothesis instead of one- or non-directional hypothesis (ablation) | MEMaBiD · test power vs sample size (Figure 5) | test power | Improved |
| Vary adaptive parameter beta from 1.0 to 2.0 | MEMaBiD · Type-I error (Figure 7) | type-I error | No clear effect |
| Vary number of local regions s | Our method · FWER (Figure 8) | familywise error rate | No clear effect |
Closest papers by meaning
- Conditional Testing based on Localized Conformal $p$-values (ICLR 2025 · similarity 0.86)
- Kernel-based Maximum-of-difference Test for Two-sample Comparison (ICML 2026 · similarity 0.85)
- Kernel Selection is Model Selection: A Unified Complexity-Penalised Approach for MMD Two-Sample Tests (NeurIPS 2026 · similarity 0.85)
- Network two-sample test for block models (NeurIPS 2025 · similarity 0.84)
- DUAL: Learning Diverse Kernels for Aggregated Two-sample and Independence Testing (NeurIPS 2025 · similarity 0.83)
- Minimax Optimal Two-Sample Testing under Local Differential Privacy (NeurIPS 2026 · similarity 0.82)
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.
