Research map / Learning theory and algorithms / Kernels and regression theory
On the Generalization Power of Overfitted Two-Layer Neural Tangent Kernel Models
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 |
|---|---|---|---|
| Use ground-truth function in learnable class F_l2 (frequencies k in {0,1,2,4}) instead of non-learnable class (k in {3,5,7,9}) | min-l2 NTK overfitted model · d=2, n=50, p large (Fig. 1) | test MSE vs null risk | Helped |
| Increase number of neurons p for NTK model with noise (ablation) | min-l2 NTK overfitted model · Fixed n=50, sigma^2 in {0,0.04,0.16} (Fig. 3a) | test MSE | No clear effect |
| Increase training samples n with noise (sigma^2=0.01) at fixed p=20000 (ablation) | min-l2 NTK overfitted model · Fig. 3(b) | test MSE gap due to noise | No clear effect |
| Increase p for min-l2 Fourier-feature linear model (vs NTK) | min-l2 Fourier features · Fig. 1(a) learnable ground truth | test MSE | Hurt |
Closest papers by meaning
- On the Generalization Power of the Overfitted Three-Layer Neural Tangent Kernel Model (NeurIPS 2022 · similarity 0.93)
- Stable Minima Cannot Overfit in Univariate ReLU Networks: Generalization by Large Step Sizes (NeurIPS 2024 · similarity 0.89)
- Generalization in Kernel Regression Under Realistic Assumptions (ICML 2024 · similarity 0.88)
- Simplicity Bias and Optimization Threshold in Two-Layer ReLU Networks (ICML 2025 · similarity 0.88)
- Benign Overfitting in Adversarial Training of Neural Networks (ICML 2024 · similarity 0.87)
- Stable Minima of ReLU Neural Networks Suffer from the Curse of Dimensionality: The Neural Shattering Phenomenon (NeurIPS 2025 · similarity 0.87)
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.