Research map / Optimization / Implicit bias and generalization
On feature learning in neural networks with global convergence guarantees
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 a partially trained three-layer network with the proposed scaling instead of NTK scaling | Partially trained three-layer neural network · Experiment 3 | test error | Helped |
| Use a partially trained three-layer network with the proposed scaling instead of NTK or MF scaling | Partially trained three-layer neural network · Experiment 3 | feature learning | Helped |
| Add a quadratic input embedding | Shallow neural network · Experiment 2 | test error | Helped |
| Train the second-to-last layer of a wide multi-layer network using gradient flow | Partially trained multi-layer neural network · Training data with unrestricted input dimension | training loss convergence | Helped |
| Use mean-field scaling for a shallow network when the training inputs are linearly independent | Shallow neural network · Training data with n ≤ d | training loss convergence | Helped |
Closest papers by meaning
- Global Convergence and Rich Feature Learning in $L$-Layer Infinite-Width Neural Networks under $\mu$ Parametrization (ICML 2025 · similarity 0.91)
- An Improved Analysis of Training Over-parameterized Deep Neural Networks (NeurIPS 2019 · similarity 0.90)
- Fast Convergence of Natural Gradient Descent for Over-Parameterized Neural Networks (NeurIPS 2019 · similarity 0.90)
- On the Surprising Effectiveness of Large Learning Rates under Standard Width Scaling (NeurIPS 2025 · similarity 0.88)
- How does Gradient Descent Learn Features --- A Local Analysis for Regularized Two-Layer Neural Networks (NeurIPS 2024 · similarity 0.88)
- Convergence Analysis of Newton's Method for Neural Networks in the Overparameterized Limit (NeurIPS 2026 · 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.