Research map / Learning theory and algorithms / Kernels and regression theory
Neural Tangent Kernel Maximum Mean Discrepancy
Xiuyuan Cheng, Yao Xie
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 |
|---|---|---|---|
| Use NTK-MMD instead of the strongest listed baseline, C2ST-L with Adam for 10 epochs | two-sample testing · Gaussian mixture Example 1, d=10 | test power: 14.1 (87.8) -> 7.1; 38.4 (100.0) -> 9.6; 76.4 (100.0) -> 13.7; 92.9 (100.0) -> 17.9 | Hurt |
| Use NTK-MMD instead of C2ST-L with Adam for 1 epoch | two-sample testing · Gaussian mixture Example 2, d=10 | test power: 48.5 -> 34.3; 92.8 -> 68.9; 99.5 -> 88.8; 100.0 -> 95.9 | Mixed |
| Use NTK-MMD instead of C2ST-S with SGD for 1 epoch | two-sample testing · Gaussian mixture Example 1, d=10 | test power: 6.0 -> 7.1; 10.6 -> 9.6; 10.8 -> 13.7; 14.8 -> 17.9 | Improved |
| Use NTK-MMD instead of C2ST-S with SGD or Adam for 1 epoch | two-sample testing · Gaussian mixture Example 2, d=10 | test power: NTK-MMD: 34.3, 68.9, 88.8, 95.9; C2ST-S (Adam): 21.8, 62.2, 79.4, 94.6; C2ST-S (SGD): 7.4, 22.7, 35.3, 54.9 | Improved |
| Increase the number of training samples processed during online training | NTK-MMD · Gaussian covariance shift | testing power | Improved |
| Replace the full NTK-MMD computation with test-only bootstrap calibration | asymmetric NTK-MMD · Gaussian covariance shift | testing power | No clear effect |
| Use NTK-MMD instead of Gaussian MMD or Hotelling's T-test | two-sample testing · Gaussian mean shift in R^100 | testing power | Mixed |
| Use NTK-MMD instead of Hotelling's T-test | change-point detection · MSRC-12 human activity sequence | detection statistic | Improved |
Closest papers by meaning
- Kernel-based Maximum-of-difference Test for Two-sample Comparison (ICML 2026 · similarity 0.87)
- Extrapolation and Spectral Bias of Neural Nets with Hadamard Product: a Polynomial Net Study (NeurIPS 2022 · similarity 0.87)
- Optimized Random Features for the Neural Tangent Kernel (Student Abstract) (AAAI 2025 · similarity 0.87)
- Training-Free Determination of Network Width via Neural Tangent Kernel (ICLR 2026 · similarity 0.87)
- The Surprising Effectiveness of Infinite-Width NTKs for Characterizing and Improving Model Training (AAAI 2025 · similarity 0.86)
- Kernel Selection is Model Selection: A Unified Complexity-Penalised Approach for MMD Two-Sample Tests (NeurIPS 2026 · similarity 0.85)
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.
