Research map / Optimization / Optimizers
Optimal Learning for Multi-pass Stochastic Gradient Methods
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 mini-batch size b = sqrt(m) instead of b = 1 (keeping optimal rate) allowing larger step-size eta_t ~ 1/sqrt(m) versus 1/m | multi-pass SGM for least squares · theoretical analysis (Corollary 3.1 / 3.6 vs 3.4) | excess risk convergence rate | Helped |
| Early stopping of multi-pass SGM with a universal step-size (number of passes as regularization parameter) | SGM with fixed step-size eta ~ 1/m · theoretical analysis (Corollary 3.4) | excess risk convergence rate | Helped |
| Simulation: mini-batch SGM, simple SGM and batch GM perform similarly on BreastCancer classification error | mini-batch SGM vs SGM vs batch GM · BreastCancer | classification error | No clear effect |
| Simulation: computational variance negligible relative to bias and sample error in synthetic experiments | mini-batch SGM and SGM · synthetic regression, m=100 | error decomposition (computational variance) | No clear effect |
| Larger number of passes increases sample error while decreasing bias (early stopping trade-off) | SGM with multiple passes · synthetic regression, m=100 | bias and sample error vs passes | Mixed |
Closest papers by meaning
- Generalization Properties and Implicit Regularization for Multiple Passes SGM (ICML 2016 · similarity 0.90)
- An Even More Optimal Stochastic Optimization Algorithm: Minibatching and Interpolation Learning (NeurIPS 2021 · similarity 0.89)
- Statistical Optimality of Stochastic Gradient Descent on Hard Learning Problems through Multiple Passes (NeurIPS 2018 · similarity 0.89)
- Hybrid Stochastic-Deterministic Minibatch Proximal Gradient: Less-Than-Single-Pass Optimization with Nearly Optimal Generalization (ICML 2020 · similarity 0.88)
- The Sample Complexity of Gradient Descent in Stochastic Convex Optimization (NeurIPS 2024 · similarity 0.88)
- Rapid Overfitting of Multi-Pass SGD in Stochastic Convex Optimization (ICML 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.