Research map / Learning theory and algorithms / Kernels and regression theory
The Neural Tangent Kernel in High Dimensions: Triple Descent and a Multi-Scale Theory of Generalization
Ben Adlam, Jeffrey Pennington
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 |
|---|---|---|---|
| Vary model parameterization across the linear and quadratic scaling transitions | NTK regression · Gaussian inputs and wide-teacher targets | test error | Mixed |
| Add the second-layer-weight kernel component K2 | Full NTK · Quadratic overparameterization transition | test error non-monotonicity | Hurt |
| Increase the number of training examples through the n0=m or n1=m vicinity | NTK regression at low SNR · Learning curves | test error | Hurt |
| Use the superabundant-width limit | NTK regression with finite SNR · Large-width learning curves | test error learning-curve rate | Improved |
| Use the superabundant-width limit in the noiseless case | NTK regression · Large-width learning curves | test error learning-curve rate | Improved |
| Remove centering (ablation) | NTK regression · Large-width regime | test error | Mixed |
| Reduce σW2 so the kernel approaches K2 (ablation) | Full NTK · Quadratic overparameterization transition | test error hump | Hurt |
Closest papers by meaning
- On the Role of Optimization in Double Descent: A Least Squares Study (NeurIPS 2021 · similarity 0.91)
- Double Trouble in Double Descent: Bias and Variance(s) in the Lazy Regime (ICML 2020 · similarity 0.90)
- On the Generalization Power of Overfitted Two-Layer Neural Tangent Kernel Models (ICML 2021 · similarity 0.90)
- Generalization in Kernel Regression Under Realistic Assumptions (ICML 2024 · similarity 0.89)
- Double-Descent Curves in Neural Networks: A New Perspective Using Gaussian Processes (AAAI 2024 · similarity 0.89)
- On the Impacts of the Random Initialization in the Neural Tangent Kernel Theory (NeurIPS 2024 · similarity 0.88)
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.
