Research map / Bandits and online learning / Multi-armed bandits
Pure Exploration and Regret Minimization in Matching 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 |
|---|---|---|---|
| Exploit the rank-1 adjacency-matrix assumption | Matching bandit algorithms · Pair and full matching selection | sample complexity and regret dependence on number of vertices: up to reaching a linear dependency in the number of vertices (up to polylog terms) | Helped |
| Use Pair-Elim with different elimination timescales for rows and columns | Bipartite pair selection · Regret minimization | regret: c_u A(u,v) log(T) + c_p log(log(T)) | Helped |
| Use Pair-Elim to identify the best pair under rank-1 structure | Bipartite pair selection · Pure exploration | sample complexity: O((N+M)log(1/δ)) | Helped |
| Compare items against the highest-parameter available item | Pair selection algorithms · Pair selection | samples needed to compare items | Helped |
| Use Pair-Elim-Mono for monopartite pair selection | Monopartite pair selection · Regret minimization | regret: c_u A(u) log(T) + c_p log log(T) | Helped |
| Use Pair-Elim-Mono for monopartite pair selection | Monopartite pair selection · Pure exploration | sample complexity: c_u A(u) log(1/δ)+c_p log log(1/δ) | Helped |
| Use Matching-id to rank items by comparing unranked items against candidate top items | Matching selection · Pure exploration | sample complexity: c_u (1/γ_min^2) log(1/δ) + c_p | Helped |
| Use Adaptive-Matching's divide-and-conquer cluster ranking | Matching selection · Regret minimization | regret: c_u N log(N)/Δ_min log(T) + c_p | Helped |
| Use Pair-Elim instead of Rank1Elim | Bipartite pair selection · Numerical problem instances described in the paper | cumulative regret: nearly a 1/3 reduction | Helped |
| Use simple-Adaptive-Matching instead of ESCB | Matching selection with equal-parameter matched items · Numerical problem instances described in the paper | regret | Mixed |
| Use Adaptive-Matching rather than an exploration-first strategy | Matching selection · Regret before unranked items are partitioned into ranked clusters | regret: U_D/U_I≤c_uN; lim_{u_2/u_1 → 0} R_D/R_I = +∞ | Mixed |
Closest papers by meaning
- Stochastic Matching Bandits with Rare Optimization Updates (NeurIPS 2026 · similarity 0.91)
- Thompson Sampling For Combinatorial Bandits: Polynomial Regret and Mismatched Sampling Paradox (NeurIPS 2024 · similarity 0.88)
- Improved Analysis for Bandit Learning in Matching Markets (NeurIPS 2024 · similarity 0.88)
- Active Ranking and Matchmaking, with Perfect Matchings (ICML 2024 · similarity 0.87)
- Optimal and Efficient Contextual Combinatorial Semi-bandits with General Function Approximation (NeurIPS 2026 · similarity 0.87)
- Adversarial Combinatorial Semi-bandits with Graph Feedback (ICML 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.