Research map / Optimization / Implicit bias and generalization
Good regularity creates large learning rate implicit biases: edge of stability, balancing, and catapult
Yuqing Wang, Zhenghao Xu, Tuo Zhao, Molei Tao
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 |
|---|---|---|---|
| use objective with degree of regularity greater than 1 (bad regularity) | GD with large learning rate · neural network training (CIFAR-10) | edge of stability occurrence: dor>1: no EoS | Hurt |
| use objective with degree of regularity greater than 2 | GD with large learning rate · neural network training (CIFAR-10) | balancing occurrence: dor>2: no balancing | Hurt |
| add batch normalization to bad-regularity networks (ablation) | GD with large learning rate · neural network training (CIFAR-10) | edge of stability occurrence: EoS reappears in all three bad-regularity cases | Improved |
| use small instead of large learning rate (ablation) | GD · neural network training (CIFAR-10) | large learning rate phenomena: phenomena vanish | Hurt |
| switch dataset from CIFAR-10 to MNIST (ablation) | GD with large learning rate · neural network training | edge of stability and balancing | No clear effect |
Closest papers by meaning
- Conflicting Biases at the Edge of Stability: Norm versus Sharpness Regularization (ICML 2026 · similarity 0.90)
- Large Learning Rate Tames Homogeneity: Convergence and Balancing Effect (ICLR 2022 · similarity 0.89)
- Flavors of Margin: Implicit Bias of Steepest Descent in Homogeneous Neural Networks (ICLR 2025 · similarity 0.87)
- Non-asymptotic Analysis of Biased Adaptive Stochastic Approximation (NeurIPS 2024 · similarity 0.87)
- Implicit Bias of Gradient Descent for Non-Homogeneous Deep Networks (ICML 2025 · similarity 0.87)
- Convergence Rates for Gradient Descent on the Edge of Stability for Overparametrised Least Squares (NeurIPS 2025 · similarity 0.86)
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.
