Research map / Optimization / Nonconvex and smooth optimization
Dynamics of Stochastic Momentum Methods on Large-scale, Quadratic Models
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 dimension-adjusted Nesterov momentum (SDANA) instead of SGD | Stochastic optimization on large-scale least-squares problems · Non-strongly convex, average-case | convergence rate: $Rt^{-3/2} + \widetilde{R} t^{-1/2}$ -> $Rt^{-3} + \widetilde{R} t^{-1}$ | Helped |
| Use dimension-adjusted Nesterov momentum (SDANA) instead of SGD | Stochastic optimization on large-scale least-squares problems · Strongly convex, average-case | convergence rate | No clear effect |
| Use fixed-parameter stochastic heavy-ball momentum (SHB) instead of SGD, with the corresponding learning-rate identification | Large-scale least-squares problems · Large-sample limit | loss dynamics | No clear effect |
| Use dimension-adjusted heavy-ball (SDAHB) instead of tuned SGD | Stochastic optimization on large-scale least-squares problems · Strongly convex | convergence rate: never more than a factor of 2 faster | Helped |
| Use dimension-adjusted heavy-ball (SDAHB) instead of SGD | Stochastic optimization on large-scale least-squares problems · Non-strongly convex | average-case convergence rate | No clear effect |
Closest papers by meaning
- Dimension-adapted Momentum Outscales SGD (NeurIPS 2025 · similarity 0.92)
- Perfect Parallelization in Mini-Batch SGD with Classical Momentum Acceleration (NeurIPS 2026 · similarity 0.92)
- Accelerated Convergence of Stochastic Heavy Ball Method under Anisotropic Gradient Noise (ICLR 2024 · similarity 0.89)
- Gradient correlation is a key ingredient to accelerate SGD with momentum (ICLR 2025 · similarity 0.89)
- Stochastic Polyak Step-sizes and Momentum: Convergence Guarantees and Practical Performance (ICLR 2025 · similarity 0.88)
- Random Scaling and Momentum for Non-smooth Non-convex Optimization (ICML 2024 · 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.