Research map / Bandits and online learning / Online convex optimization and games
MetaGrad: Multiple Learning Rates in Online Learning
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 |
|---|---|---|---|
| Replace fixed decreasing learning rates with MetaGrad's master that aggregates multiple learning rates via tilted exponential weights | online convex optimization (MetaGrad full version) · 1-D simulation: fixed f(u)=|u-1/4| and stochastic |u-x_t| | regret growth rate | Helped |
| Use the Bernstein-condition / directional-derivative curvature conditions to get logarithmic regret for functions without curvature (e.g., unregularized hinge loss) | online convex optimization with hinge loss · i.i.d. hinge loss on unit ball with beta=1 | regret rate | Helped |
Closest papers by meaning
- Dual Adaptivity: A Universal Algorithm for Minimizing the Adaptive Regret of Convex Functions (NeurIPS 2021 · similarity 0.91)
- Adaptive Gradient-Based Meta-Learning Methods (NeurIPS 2019 · similarity 0.90)
- A Simple yet Universal Strategy for Online Convex Optimization (ICML 2022 · similarity 0.89)
- Learning-to-learn non-convex piecewise-Lipschitz functions (NeurIPS 2021 · similarity 0.88)
- Small-loss Adaptive Regret for Online Convex Optimization (ICML 2024 · similarity 0.88)
- Provable and Practical Online Learning Rate Adaptation with Hypergradient Descent (ICML 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.