Research map / Bandits and online learning / Multi-armed bandits
An Asymptotically Optimal Batched Algorithm for the Dueling Bandit Problem
Arpit Agarwal, Rohan Ghuge, Viswanath Nagarajan
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 C2B with at most log(T) batches instead of fully sequential updates | K-armed dueling bandits under the Condorcet condition · Real-world preference datasets | cumulative regret | No clear effect |
| Use C2B with B = floor(log(T)) batches instead of BTM | Dueling-bandit experiments · Six rankers, Sushi, Irish-Meath, Irish-Dublin, MSLR30, Yahoo30 | cumulative regret | Improved |
| Use C2B with B = floor(log(T)) batches instead of SCOMP2 | Dueling-bandit experiments · Most tested datasets | cumulative regret | Improved |
| Use C2B with B = floor(log(T)) batches instead of RUCB | Dueling-bandit experiments · All datasets except Yahoo30 | cumulative regret | No clear effect |
| Use C2B with B = floor(log(T)) + 6 batches instead of RUCB | Dueling-bandit experiments · Six rankers, Sushi, Irish-Meath, Irish-Dublin, MSLR30, Yahoo30 | cumulative regret | Improved |
| Use C2B with B = floor(log(T)) + 6 batches instead of RMED | Dueling-bandit experiments · All datasets except Yahoo30 | cumulative regret | No clear effect |
| Increase C2B's batch count by a constant number of additional rounds (ablation) | C2B · Real-world preference datasets | cumulative regret | Improved |
| Replace the elimination criterion with KL-divergence-based elimination (ablation) | C2B · Computational experiments on real-world datasets | cumulative regret | Improved |
| Use a random seed-set selection strategy | SCOMP2 · Repeated runs on the same dataset | cumulative regret | Mixed |
Closest papers by meaning
- On Weak Regret Analysis for Dueling Bandits (NeurIPS 2024 · similarity 0.93)
- Variance-aware Regret Bounds for Stochastic Contextual Dueling Bandits (ICLR 2024 · similarity 0.91)
- Borda Regret Minimization for Generalized Linear Dueling Bandits (ICML 2024 · similarity 0.91)
- Neural Dueling Bandits: Preference-Based Optimization with Human Feedback (ICLR 2025 · similarity 0.90)
- Adaptive Regret for Bandits Made Possible: Two Queries Suffice (ICLR 2024 · similarity 0.90)
- Lipschitz Dueling Bandits over Continuous Action Spaces (NeurIPS 2026 · 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.
