Research map / Point clouds / Point cloud classification
Efficient Change-Point Detection for Tackling Piecewise-Stationary Bandits
Lilian Besson, Emilie Kaufmann, Odalric-Ambrym Maillard, Julien Seznec
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 |
|---|---|---|---|
| Reduce forced exploration by setting α₀ from 1 to 0.5 (ablation) | GLR-klUCB Global · Random piecewise-stationary instances (T=20000) | regret: 611 -> 404 | Improved |
| Reduce forced exploration by setting α₀ from 0.5 to 0.1 (ablation) | GLR-klUCB Global · Random piecewise-stationary instances (T=20000) | regret: 404 -> 252 | Improved |
| Reduce forced exploration by setting α₀ from 1 to 0.5 (ablation) | GLR-klUCB Local · Random piecewise-stationary instances (T=20000) | regret: 691 -> 462 | Improved |
| Reduce forced exploration by setting α₀ from 0.5 to 0.1 (ablation) | GLR-klUCB Local · Random piecewise-stationary instances (T=20000) | regret: 462 -> 284 | Improved |
| Use Global rather than Local restarts (ablation) | GLR-klUCB · Random piecewise-stationary instances (T=20000) | regret: 691 -> 611; 462 -> 404; 284 -> 252 | Improved |
| Reduce forced exploration to α₀=0.5 (ablation) | GLR-klUCB versus klUCB · Random piecewise-stationary instances (T=20000) | regret: 515 -> 404 | Improved |
| Reduce forced exploration to α₀=0.1 (ablation) | GLR-klUCB versus klUCB · Random piecewise-stationary instances (T=20000) | regret: 515 -> 252 | Improved |
| Use GLR-klUCB Global instead of non-stationary baselines | Random piecewise-stationary bandit algorithms · Random piecewise-stationary instances (T=20000), α₀=1 | regret: 826 -> 611; 1022 -> 611; 1083 -> 611; 1070 -> 611; 973 -> 611 | Improved |
| Use GLR-klUCB Global instead of klUCB | Random piecewise-stationary bandit algorithms · Random piecewise-stationary instances (T=20000), α₀=1 | regret: 515 -> 611 | Hurt |
| Use the time-varying exploration sequence αₖ instead of constant exploration α (ablation) | GLR-klUCB · Problems 1 and 2 | regret | Improved |
| Use Global rather than Local restarts (ablation) | GLR-klUCB · Problems 1 and 2 | regret | No clear effect |
| Use GLR-klUCB instead of AdSwitch | Piecewise-stationary bandit algorithms · Problem 1, Problem 2, T=5000 | regret | Improved |
Closest papers by meaning
- Fixed-Confidence Multiple Change Point Identification under Bandit Feedback (ICML 2025 · similarity 0.86)
- The Sample Complexity of Multiple Change Point Identification under Bandit Feedback (NeurIPS 2026 · similarity 0.86)
- Almost Minimax Optimal Best Arm Identification in Piecewise Stationary Linear Bandits (NeurIPS 2024 · similarity 0.85)
- Non-Stationary Lipschitz Bandits (NeurIPS 2025 · similarity 0.84)
- Adaptive Regret for Bandits Made Possible: Two Queries Suffice (ICLR 2024 · similarity 0.81)
- Constrained Feedback Learning for Non-Stationary Multi-Armed Bandits (NeurIPS 2025 · similarity 0.81)
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.
