Research map / Optimization / Implicit bias and generalization
Model, sample, and epoch-wise descents: exact solution of gradient flow in the random feature model
Antoine Bodin, Nicolas Macris
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 |
|---|---|---|---|
| stop gradient-flow training at an earlier time rather than using the minimum least-squares estimator | random feature model near the interpolation threshold · model-wise double descent | test error | Improved |
| vary the number of model parameters across the interpolation threshold | random feature model during gradient flow · model-wise double descent | test error | Mixed |
| vary the number of training samples across the interpolation threshold | random feature model during gradient flow · sample-wise double descent | test error | Mixed |
| train over time in the largely overparameterized regime | random feature model with ψ ≫ φ · epoch-wise descent | test error | Mixed |
| use small regularization in the described noisy, overparameterized setting | random feature model with ψ = 2φ · epoch-wise double descent | test error | Mixed |
| vary sample size and model size so both interpolation conditions are encountered | random feature model at large training time · triple descent | test error | Mixed |
Closest papers by meaning
- On the Role of Optimization in Double Descent: A Least Squares Study (NeurIPS 2021 · similarity 0.91)
- The Neural Tangent Kernel in High Dimensions: Triple Descent and a Multi-Scale Theory of Generalization (ICML 2020 · similarity 0.90)
- Early Stopping in Deep Networks: Double Descent and How to Eliminate it (ICLR 2021 · similarity 0.89)
- Unraveling the Enigma of Double Descent: An In-depth Analysis through the Lens of Learned Feature Space (ICLR 2024 · similarity 0.89)
- Double-Descent Curves in Neural Networks: A New Perspective Using Gaussian Processes (AAAI 2024 · similarity 0.88)
- Deep Linear Network Training Dynamics from Random Initialization: Data, Width, Depth, and Hyperparameter Transfer (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.
