Research map / Learning theory and algorithms / Kernels and regression theory
Understanding Double Descent Requires A Fine-Grained Bias-Variance Decomposition
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 |
|---|---|---|---|
| Increase random-feature width | Random feature kernel regression · High-dimensional random feature kernel regression | bias | Helped |
| Increase random-feature width toward the interpolation boundary | Random feature kernel regression · Interpolation boundary | variance component V_PX | Hurt |
| Increase random-feature width toward the interpolation boundary | Random feature kernel regression · Interpolation boundary | variance component V_PXε | Hurt |
| Add sampling and initialization interaction variance | Random feature kernel regression · Interpolation boundary, including the noiseless setting | test loss | Hurt |
| Add label noise | Random feature kernel regression · Interpolation boundary | test loss | Hurt |
| Average independently initialized learners | Random feature kernel regression · Interpolation boundary | test loss | Helped |
| Average learners trained on independent data samples | Random feature kernel regression · Interpolation boundary | test loss | Helped |
| Retain the initial function value instead of centering the predictor (ablation) | Full NTK regression · High-dimensional NTK regression | variance | Hurt |
| Remove the initial function value by centering the predictor (ablation) | Full NTK regression · High-dimensional NTK regression | variance | Helped |
| Remove the first-layer NTK contribution, yielding the standard random-features kernel (ablation) | Full NTK regression · High-dimensional NTK regression | test loss | Mixed |
Closest papers by meaning
- Unraveling the Enigma of Double Descent: An In-depth Analysis through the Lens of Learned Feature Space (ICLR 2024 · similarity 0.89)
- Simplicity Bias in Overparameterized Machine Learning (AAAI 2024 · similarity 0.88)
- Bias of Stochastic Gradient Descent or the Architecture: Disentangling the Effects of Overparameterization of Neural Networks (ICML 2024 · similarity 0.87)
- Simplicity Bias and Optimization Threshold in Two-Layer ReLU Networks (ICML 2025 · similarity 0.87)
- Subsampling is not Magic: Why Large Batch Sizes Work for Differentially Private Stochastic Optimisation (ICML 2024 · similarity 0.87)
- Position: Deep Learning is Not So Mysterious or Different (ICML 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.