Research map / Bandits and online learning / Multi-armed bandits
Non-Stationary Bandits with Auto-Regressive Temporal Dependency
Qinyi Chen, Negin Golrezaei, Djallel Bouneffouf
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 AR2 instead of epsilon-greedy | epsilon-greedy · Tourism case study | normalized regret: 0.38 (0.16) -> 0.26 (0.14) | Improved |
| use AR2 instead of modified UCB | UCB-mod · Tourism case study | normalized regret: 0.60 (0.27) -> 0.26 (0.14) | Improved |
| use AR2 instead of epsilon-greedy | epsilon-greedy · Tourism case study | optimal arm pulls: 133.83 (12.47) -> 142.04 (15.71) | Improved |
| use AR2 instead of epsilon-greedy | epsilon-greedy · Synthetic E[alpha]=0.9, k=2 | normalized regret: 0.36 (0.14) -> 0.18 (0.06) | Improved |
| use AR2 instead of mod-UCB | mod-UCB · Synthetic E[alpha]=0.9, k=20 | normalized regret: 0.53 (0.03) -> 0.49 (0.02) | Improved |
| use AR2 instead of mod-UCB | mod-UCB · Synthetic E[alpha]=0.4, k=10 | normalized regret: 0.67 (0.03) -> 0.67 (0.01) | No clear effect |
| use AR2 instead of predictive sampling | PS · Synthetic E[alpha]=0.9, k=2 | normalized regret: 0.52 (0.33) -> 0.18 (0.06) | Improved |
| use AR2 instead of UCB | UCB · Synthetic E[alpha]=0.4, k=2 | normalized regret: 0.68 (0.09) -> 0.38 (0.10) | Improved |
| use AR2 instead of explore-then-commit | ETC · Synthetic E[alpha]=0.4, k=2 | normalized regret: 1.00 (0.03) -> 0.38 (0.10) | Improved |
| use AR2 instead of sliding-window UCB | SW-UCB · Synthetic E[alpha]=0.9, k=2 | normalized regret: 0.55 (0.13) -> 0.18 (0.06) | Improved |
| increase number of arms from 2 to 20 | AR2 · Synthetic E[alpha]=0.9 | normalized regret: 0.18 (0.06) -> 0.49 (0.02) | Hurt |
| larger AR parameter (0.4 to 0.9) | AR2 · Synthetic k=2 | normalized regret: 0.38 (0.10) -> 0.18 (0.06) | Improved |
Closest papers by meaning
- Stochastic Rising Bandits (ICML 2022 · similarity 0.90)
- Tightening Regret Lower and Upper Bounds in Restless Rising Bandits (NeurIPS 2025 · similarity 0.89)
- Constrained Feedback Learning for Non-Stationary Multi-Armed Bandits (NeurIPS 2025 · similarity 0.89)
- Forced Exploration in Bandit Problems (AAAI 2024 · similarity 0.87)
- Non-stationary Bandit Convex Optimization: A Comprehensive Study (NeurIPS 2025 · similarity 0.87)
- Block Optimism for Nonstationary Bandits with Latent Linear Dynamics (NeurIPS 2026 · 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.
