Research map / Optimization / Implicit bias and generalization
Fine-Grained Analysis of Optimization and Generalization for Overparameterized Two-Layer Neural Networks
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 true labels instead of random labels | Two-layer ReLU network trained by gradient descent · MNIST | training convergence rate | Helped |
| Use true labels instead of random labels | Two-layer ReLU network trained by gradient descent · CIFAR | training convergence rate | Helped |
| Align labels with the eigenvector corresponding to the least eigenvalue | Two-layer ReLU network trained by gradient descent · MNIST and CIFAR | training convergence rate | Hurt |
| Increase the portion of randomly assigned labels | Data-dependent complexity measure for the learned network · MNIST and CIFAR | data-dependent complexity measure | Hurt |
| Increase the portion of randomly assigned labels | Two-layer ReLU network trained by gradient descent · MNIST and CIFAR | generalization error | Hurt |
Closest papers by meaning
- Simplicity Bias and Optimization Threshold in Two-Layer ReLU Networks (ICML 2025 · similarity 0.92)
- Generalization of Two-layer Neural Networks: An Asymptotic Viewpoint (ICLR 2020 · similarity 0.91)
- Stable Minima Cannot Overfit in Univariate ReLU Networks: Generalization by Large Step Sizes (NeurIPS 2024 · similarity 0.90)
- Exact, Tractable Gauss-Newton Optimization in Deep Reversible Architectures Reveal Poor Generalization (NeurIPS 2024 · similarity 0.89)
- How Uniform Random Weights Induce Non-uniform Bias: Typical Interpolating Neural Networks Generalize with Narrow Teachers (ICML 2024 · similarity 0.89)
- Benign Overfitting and Grokking in ReLU Networks for XOR Cluster Data (ICLR 2024 · similarity 0.89)
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.