Research map / Bandits and online learning / Contextual bandits and Thompson sampling
Reward Imputation with Sketching for Contextual Batched Bandits
Xiao Zhang, Ninglu Shao, Zihua Si, Jun Xu, Wenhan Wang, Hanjing Su, Ji-Rong Wen
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 |
|---|---|---|---|
| reward imputation (SPUIR) vs second-best baseline | contextual batched bandit policy · commercial product | CVR: 1.07% | Improved |
| reward imputation (SPUIR) vs second-best baseline | contextual batched bandit policy · commercial product | CTCVR: 1.12% | Improved |
| add reward imputation to batched bandit (PUIR vs SBUCB) | SBUCB · commercial product | CVR: 0.8569 -> 0.8807 | Improved |
| use sketching in reward imputation (SPUIR vs PUIR) | PUIR · commercial product | time (sec): 184.3575 -> 81.5753 | Improved |
| use sketching in reward imputation (SPUIR vs PUIR) | PUIR · commercial product | CVR: 0.8807 -> 0.8770 | No clear effect |
| add rate scheduling to sketched imputation (SPUIR-RS vs SPUIR) | SPUIR · commercial product | CVR: 0.8770 -> 0.8758 | No clear effect |
| add rate scheduling to non-sketched imputation (PUIR-RS vs PUIR) | PUIR · commercial product | CTCVR: 0.4411 -> 0.4389 | No clear effect |
| sketched imputation vs DFM-S baseline runtime | SPUIR · commercial product | time (sec): 302.3140 -> 81.5753 | Improved |
| sketched imputation vs BLTS-B baseline runtime | SPUIR · commercial product | time (sec): 218.2109 -> 81.5753 | Improved |
| sketching cost versus no-imputation baselines | SPUIR · commercial product | time (sec): 43.5435 -> 81.5753 | Hurt |
| use nonlinear kernel reward model (SPUIR-Kernel) | SPUIR · synthetic dataset | average reward | Improved |
| batch size B set to 1,000 (C_B = 25) versus other sizes (ablation) | SPUIR · Criteo-recent | average reward | Improved |
Closest papers by meaning
- Active Context Selection Improves Simple Regret in Contextual Bandits (NeurIPS 2026 · similarity 0.87)
- Mixed-Effects Contextual Bandits (AAAI 2024 · similarity 0.87)
- Feel-Good Thompson Sampling for Contextual Bandits: a Markov Chain Monte Carlo Showdown (NeurIPS 2025 · similarity 0.87)
- Factored-Reward Bandits with Intermediate Observations (ICML 2024 · similarity 0.87)
- Causal Bandits: The Pareto Optimal Frontier of Adaptivity, a Reduction to Linear Bandits, and Limitations around Unknown Marginals (ICML 2024 · similarity 0.87)
- Corruption-Robust Linear Bandits: Minimax Optimality and Gap-Dependent Misspecification (NeurIPS 2024 · 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.
