Research map / Bandits and online learning / Multi-armed bandits
Coordinate Descent with Bandit Sampling
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 |
|---|---|---|---|
| select the coordinate with the largest marginal-decrease lower bound | coordinate descent (max_r) · aloi / Lasso | time to sub-optimality gap exp(-5) (seconds): 27.8 -> 6.2 | Helped |
| select the coordinate with the largest marginal-decrease lower bound | coordinate descent (max_r) · a9a / logistic regression | time to sub-optimality gap exp(-5) (seconds): 11.8 -> 4.5 | Helped |
| select the coordinate with the largest marginal-decrease lower bound | coordinate descent (max_r) · usps / ridge regression | time to sub-optimality gap exp(-5) (seconds): 1 -> 9.5 | Hurt |
| use bandit estimates with periodic full-coordinate refresh and exploration | coordinate descent (B_max_r) · aloi / Lasso | time to sub-optimality gap exp(-5) (seconds): 27.8 -> 11 | Helped |
| use bandit estimates with periodic full-coordinate refresh and exploration | coordinate descent (B_max_r) · a9a / logistic regression | time to sub-optimality gap exp(-5) (seconds): 11.8 -> 1.9 | Helped |
| use bandit estimates with periodic full-coordinate refresh and exploration | coordinate descent (B_max_r) · usps / ridge regression | time to sub-optimality gap exp(-5) (seconds): 1 -> 1 | No clear effect |
| use B_max_r instead of max_r | coordinate descent · aloi / Lasso | time to sub-optimality gap exp(-5) (seconds): 6.2 -> 11 | Hurt |
| use B_max_r instead of max_r | coordinate descent · a9a / logistic regression | time to sub-optimality gap exp(-5) (seconds): 4.5 -> 1.9 | Helped |
| use B_max_r instead of max_r | coordinate descent · usps / ridge regression | time to sub-optimality gap exp(-5) (seconds): 9.5 -> 1 | Helped |
| increase the bin size E | B_max_r on a9a / logistic regression · a9a / logistic regression | iterations to log-suboptimality gap -5 | Hurt |
| increase the exploration parameter ε | B_max_r on a9a / logistic regression · a9a / logistic regression | iterations to log-suboptimality gap -5 | Hurt |
Closest papers by meaning
- Adaptive Regret for Bandits Made Possible: Two Queries Suffice (ICLR 2024 · similarity 0.87)
- Non-stationary Bandit Convex Optimization: A Comprehensive Study (NeurIPS 2025 · similarity 0.87)
- Small steps no more: Global convergence of stochastic gradient bandits for arbitrary learning rates (NeurIPS 2024 · similarity 0.87)
- Causal Bandits: The Pareto Optimal Frontier of Adaptivity, a Reduction to Linear Bandits, and Limitations around Unknown Marginals (ICML 2024 · similarity 0.87)
- Improved Regret for Bandit Convex Optimization with Delayed Feedback (NeurIPS 2024 · similarity 0.87)
- Geometry-Aware Approaches for Balancing Performance and Theoretical Guarantees in Linear Bandits (ICLR 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.