Research map / Optimization / Implicit bias and generalization
Stochasticity of Deterministic Gradient Descent: Large Learning Rate for Multiscale Objective Function
Lingkai Kong, 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 large learning rate (0.02) instead of small (0.0005) for batch gradient descent | 5-16-1 MLP with leaky ReLU · UCI Airfoil Self-Noise regression | long-time behavior of loss | Mixed |
| relax conditions on the small-scale term to a non-periodic, non-quasiperiodic function | deterministic GD with f0=x^2/2 · synthetic multiscale objective | ergodicity and mixing | No clear effect |
| replace periodic small scale with two incommensurate sinusoids | deterministic GD with f0=x^4/4 · synthetic multiscale objective | convergence to rescaled Gibbs distribution | No clear effect |
| add momentum to gradient descent | deterministic GD · synthetic multiscale objective | stochastic behavior at large learning rate | No clear effect |
Closest papers by meaning
- Noise and Fluctuation of Finite Learning Rate Stochastic Gradient Descent (ICML 2021 · similarity 0.89)
- Characterizing Dynamical Stability of Stochastic Gradient Descent in Overparameterized Learning (JMLR 2025 · similarity 0.88)
- On the Theoretical Properties of Noise Correlation in Stochastic Optimization (NeurIPS 2022 · similarity 0.88)
- High-dimensional Limit of SGD for Diagonal Linear Networks (NeurIPS 2026 · similarity 0.88)
- What is the Long-Run Distribution of Stochastic Gradient Descent? A Large Deviations Analysis (ICML 2024 · similarity 0.87)
- The Global Convergence Time of Stochastic Gradient Descent in Non-Convex Landscapes: Sharp Estimates via Large Deviations (ICML 2025 · similarity 0.87)
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.
