Research map / Bandits and online learning / Multi-armed bandits
Tight Regret Bounds for Single-pass Streaming Multi-armed Bandits
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 |
|---|---|---|---|
| Adding O(log log K) or O(log* K) arm memory eps-best arm exploration-and-commit instead of uniform exploration | single-pass streaming MAB regret minimization · Synthetic Bernoulli arms (uniform and standout streams) | relative mean regret vs uniform exploration: log log K variant reached ~0.32-0.67 of benchmark in most settings | Helped |
| Using the naive uniform elimination (single-arm) instead of uniform exploration | streaming MAB regret minimization · Synthetic Bernoulli arms, all settings | relative mean regret vs uniform exploration: relative regret mostly 1.9-3.7x benchmark | Hurt |
| Using O(log K)-arm eps-best (exponentially decreasing eps across levels) as the subroutine | streaming MAB regret minimization · standout stream, K=5000 and K=50000 at T=1000K^3 | relative mean regret vs uniform exploration: 9.12 and 20.45 relative to benchmark | Hurt |
| Using O(log K)-arm eps-best subroutine instead of uniform exploration | streaming MAB regret minimization · uniform stream, median regret | relative median regret vs uniform exploration: e.g. 0.3941 at K=500, T=1000K^3 | Helped |
| Using the O(log* K)-arm Aggressive Selective Promotion subroutine instead of uniform exploration | streaming MAB regret minimization · standout stream, K=500 and K=5000 at T=1000K^3 | relative mean regret vs uniform exploration: 0.5639 to 1.7245 across settings | Mixed |
Closest papers by meaning
- Regret Minimisation in Multi-Armed Bandits Using Bounded Arm Memory (AAAI 2020 · similarity 0.93)
- Nearly Tight Bounds for Exploration in Streaming Multi-Armed Bandits with Known Optimality Gap (AAAI 2025 · similarity 0.93)
- Linear Streaming Bandit: Regret Minimization and Fixed-Budget Epsilon-Best Arm Identification (AAAI 2025 · similarity 0.91)
- p-Mean Regret for Stochastic Bandits (AAAI 2025 · similarity 0.90)
- Improved Regret Bounds for Linear Bandits with Heavy-Tailed Rewards (NeurIPS 2025 · similarity 0.89)
- Adaptive Regret for Bandits Made Possible: Two Queries Suffice (ICLR 2024 · similarity 0.89)
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.