Research map / Optimization / Implicit bias and generalization
Implicit Bias of Gradient Descent for Logistic Regression at the Edge of Stability
Jingfeng Wu, Vladimir Braverman, Jason D. Lee
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 large stepsize eta=0.1 instead of small eta=0.01 | GD training a 4-layer FC network · MNIST 1,000-sample subset | training loss monotonicity | Hurt |
| use large stepsize eta=0.1 instead of small eta=0.01 | GD training a 4-layer FC network · MNIST 1,000-sample subset | test accuracy | Improved |
| use stepsize eta >= 0.1 instead of eta=0.01 | GD for logistic regression · MNIST digits 0 vs 8, 1,000 samples | risk monotonicity | Hurt |
| use large constant stepsize (EoS regime) instead of small stepsize | GD for logistic regression · linearly separable data (theory and simulation) | long-run risk convergence | No clear effect |
| replace logistic loss with exponential loss (ablation) | GD with large constant stepsize · linearly separable data (theory) | convergence/stability | Hurt |
Closest papers by meaning
- Gradient Descent Converges Arbitrarily Fast for Logistic Regression via Large and Adaptive Stepsizes (ICML 2025 · similarity 0.92)
- Constant Stepsize Local GD for Logistic Regression: Acceleration by Instability (ICML 2025 · similarity 0.91)
- Any-stepsize Gradient Descent for Separable Data under Fenchel–Young Losses (NeurIPS 2025 · similarity 0.91)
- Large Stepsizes Accelerate Gradient Descent for Regularized Logistic Regression (NeurIPS 2025 · similarity 0.90)
- Benefits of Early Stopping in Gradient Descent for Overparameterized Logistic Regression (ICML 2025 · similarity 0.89)
- From Logistic Regression to the Perceptron Algorithm: Exploring Gradient Descent with Large Step Sizes (AAAI 2025 · 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.
