Research map / Optimization / Optimizers
An Improved Analysis of Stochastic Gradient Descent with Momentum
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 multistage parameters (decreasing stepsize, increasing momentum) instead of fixed momentum 0.9 with the same stepsize schedule (ablation) | SGDM · CIFAR-10 ResNet-18 | training loss / validation accuracy | Helped |
| use multistage stepsizes instead of a fixed large stepsize 0.66 (ablation) | SGDM · MNIST logistic regression | final training loss | Helped |
| use multistage stepsizes instead of a fixed small stepsize 0.095 (ablation) | SGDM · MNIST logistic regression | initial convergence speed | Helped |
| increase fixed stepsize alpha from 0.1 to 0.5/1.0 | SGDM · MNIST logistic regression | training loss | Mixed |
| use momentum 0.9 instead of plain SGD at same stepsize | SGD · MNIST logistic regression | training loss | No clear effect |
Closest papers by meaning
- Random Scaling and Momentum for Non-smooth Non-convex Optimization (ICML 2024 · similarity 0.93)
- Gradient correlation is a key ingredient to accelerate SGD with momentum (ICLR 2025 · similarity 0.92)
- The Marginal Value of Momentum for Small Learning Rate SGD (ICLR 2024 · similarity 0.92)
- Better SGD using Second-order Momentum (NeurIPS 2022 · similarity 0.92)
- Dynamic Momentum Recalibration in Online Gradient Learning (CVPR 2026 · similarity 0.92)
- Derivatives of Stochastic Gradient Descent in parametric optimization (NeurIPS 2024 · similarity 0.90)
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.