Research map / Learning theory and algorithms / Sample complexity
Testing and Learning on Distributions with Symmetric Noise Invariance
Ho Chung Law, Christopher Yau, Dino Sejdinovic
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 |
|---|---|---|---|
| Test paired differences for asymmetry instead of applying ME directly to samples | Two-sample testing with differing additive symmetric noise · Synthetic noisy χ² | Type I error control | Improved |
| Apply ME directly to samples rather than using the paired-difference SME test | Two-sample testing with differing additive symmetric noise · Synthetic noisy χ² | Type I error control | Hurt |
| Use the PhD test at large additive noise levels | PhD two-sample test · Synthetic noisy χ² | Type I error control | Hurt |
| Use the PhD test rather than SME at small noise levels | Two-sample testing · Synthetic noisy χ² | power | Improved |
| Use SME rather than ME on low-level angular-momentum features | Two-sample testing · Higgs low-level features | Type I error | Improved |
| Use SME rather than ME on high-level features | Two-sample testing · Higgs high-level features, N=500 | power: 0.94 vs 1.0 | Hurt |
| Use phase features instead of Fourier embeddings | Distribution regression · Aerosol MISR1 | RMSE robustness to test noise | Improved |
| Learn discriminative frequencies with Phase NN instead of using PLRR phase features | Distribution regression · Aerosol MISR1 | RMSE robustness to covariate shift | Improved |
| Use phase features rather than Fourier embeddings | Distribution regression · Dark Matter ML1 | RMSE | Improved |
| Apply a second-level Gaussian kernel to phase features rather than approximate embeddings | Distribution regression · Dark Matter ML1 | RMSE | Mixed |
| Remove per-frequency normalization from Phase NN, yielding Fourier NN (ablation) | Phase NN · Aerosol MISR1 | RMSE robustness to covariate shift | No clear effect |
| Use MMD on paired differences rather than phase features | Learning on distributions · Synthetic noisy χ² bags | invariance to additive symmetric noise | Hurt |
Closest papers by meaning
- MMD-Fuse: Learning and Combining Kernels for Two-Sample Testing Without Data Splitting (NeurIPS 2023 · similarity 0.87)
- Learning Theory for Distribution Regression (JMLR 2016 · similarity 0.87)
- Kernel conditional tests from learning-theoretic bounds (NeurIPS 2025 · similarity 0.86)
- Revealing Distribution Discrepancy by Sampling Transfer in Unlabeled Data (NeurIPS 2024 · similarity 0.86)
- Leveraging Data Symmetries to Select an Optimal Subset of Training Data under Label Noise (NeurIPS 2026 · similarity 0.85)
- Optimal Algorithms for Augmented Testing of Discrete Distributions (NeurIPS 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.
