Research map / Optimization / Implicit bias and generalization
Self-Stabilization: The Implicit Bias of Gradient Descent at the Edge of Stability
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 gradient descent at the edge of stability, where cubic self-stabilization counteracts sharpness growth | Full-batch gradient descent · Neural-network training | sharpness | Helped |
| Use gradient descent at the edge of stability rather than unconstrained gradient flow | Gradient descent · Neural-network training | deviation from constrained trajectory | Helped |
| Use larger learning rates | Full-batch gradient descent · CIFAR10, ResNet18 | training time | Helped |
| Use larger learning rates | Full-batch gradient descent · CIFAR10, ResNet18 | generalization | Helped |
| Use generalized predicted dynamics to account for non-quadratic loss along the top eigenvector | Predicted dynamics · Mean squared error and cross-entropy experiments | dynamics prediction accuracy | Helped |
Closest papers by meaning
- Gradient Descent on Neural Networks Typically Occurs at the Edge of Stability (ICLR 2021 · similarity 0.94)
- A Minimalist Example of Edge-of-Stability and Progressive Sharpening (NeurIPS 2025 · similarity 0.91)
- Convergence Rates for Gradient Descent on the Edge of Stability for Overparametrised Least Squares (NeurIPS 2025 · similarity 0.90)
- Universal Sharpness Dynamics in Neural Network Training: Fixed Point Analysis, Edge of Stability, and Route to Chaos (ICLR 2025 · similarity 0.89)
- A Precise Characterization of SGD Stability Using Loss Surface Geometry (ICLR 2024 · similarity 0.88)
- A Unified Stability Analysis of SAM vs SGD: Role of Data Coherence and Emergence of Simplicity Bias (NeurIPS 2025 · 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.