Research map / Bandits and online learning / Contextual bandits and Thompson sampling
BFTS: Thompson Sampling with Bayesian Additive Regression Trees
Ruizhe Deng, Bibhas Chakraborty, Ran Chen, Yan Shuo Tan
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 one-hot arm encoding in a single joint model instead of separate-arm BART (ablation) | BFTS · OpenML sensitivity (4 datasets) | normalized regret: 1.000 -> 3.15 | Hurt |
| use multi encoding instead of separate-arm (ablation) | BFTS · OpenML sensitivity (4 datasets) | normalized regret: 1.051 | No clear effect |
| increase MCMC chains from 1 to 4 (ablation) | BFTS · OpenML sensitivity | normalized regret: 1.000 -> 1.071 | No clear effect |
| increase trees from 50 to 100 (ablation) | BFTS · OpenML sensitivity | normalized regret: 1.151 -> 1.058 | Improved |
| increase burn-in from 250 to 750 (ablation) | BFTS · OpenML sensitivity | normalized regret: 1.112 -> 1.079 | No clear effect |
| enable quick-decay split prior (ablation) | BFTS · OpenML sensitivity | normalized regret: 1.227 -> 1.113 | Improved |
| enable Dirichlet-sparse split prior (ablation) | BFTS · OpenML sensitivity | normalized regret: 1.154 -> 1.100 | Improved |
| use small depth prior alpha 0.1 instead of 0.3 or 0.45 (ablation) | BFTS · OpenML sensitivity | normalized regret: 1.160 | Hurt |
| vary refresh schedule (log, sqrt, hybrid) (ablation) | BFTS · OpenML sensitivity | normalized regret: 1.01 - 1.07 | No clear effect |
| use BFTS instead of best baseline LinUCB at t=1000 | BFTS · Drink Less mHealth trial | SNIPS policy value: +5.5% | Improved |
| use BFTS instead of original random policy at t=10,000 | BFTS · Drink Less mHealth trial | SNIPS policy value: +38.2% | Improved |
| use BFTS instead of LinTS tuned at t=10,000 | BFTS · Drink Less mHealth trial | SNIPS policy value: +1.5% | Improved |
Closest papers by meaning
- Feel-Good Thompson Sampling for Contextual Bandits: a Markov Chain Monte Carlo Showdown (NeurIPS 2025 · similarity 0.91)
- The Choice of Noninformative Priors for Thompson Sampling in Multiparameter Bandit Models (AAAI 2024 · similarity 0.90)
- Contextual Thompson Sampling via Generation of Missing Data (NeurIPS 2025 · similarity 0.90)
- No-Regret Thompson Sampling for Finite-Horizon Markov Decision Processes with Gaussian Processes (NeurIPS 2025 · similarity 0.90)
- Variance-Aware Feel-Good Thompson Sampling for Contextual Bandits (NeurIPS 2025 · similarity 0.89)
- Only Pay for What Is Uncertain: Variance-Adaptive Thompson Sampling (ICLR 2024 · 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.
