Research map / Bandits and online learning / Multi-armed bandits
Racing Thompson: an Efficient Algorithm for Thompson Sampling with Non-conjugate Priors
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 |
|---|---|---|---|
| Use Racing Thompson instead of standard Thompson sampling | Bernoulli bandits with conjugate Beta priors · Bernoulli bandits, 10 arms | regret | No clear effect |
| Use Racing Thompson instead of Thompson sampling with a Beta prior | Bernoulli bandits with a truncated Gaussian prior · Bernoulli bandits, 10 arms, truncated Gaussian prior | regret | Helped |
| Use Racing Thompson instead of SMC | Bernoulli bandits with a truncated Gaussian prior · Bernoulli bandits, 10 arms, truncated Gaussian prior; larger time steps | regret | Helped |
| Use Racing Thompson instead of Thompson sampling with a Beta prior | Bernoulli bandits with dependent priors · Bernoulli bandits, 10 arms, dependent prior | regret | Helped |
| Use Racing Thompson instead of SMC | Bernoulli bandits with dependent priors · Bernoulli bandits, 10 arms, dependent prior; larger time steps | regret | Helped |
| Increase σ from 0.1 to 0.7 (ablation) | Racing Thompson · Bernoulli bandits, Beta(5,5), δ=0.1 | average regret: 19.3 -> 27.8 | Hurt |
| Increase δ from 0.1 to 0.7 (ablation) | Racing Thompson · Bernoulli bandits, Beta(5,5), σ=0.1 | average regret: 19.3 -> 20.2 | Hurt |
| Decrease σ from 0.7 to 0.1 (ablation) | Racing Thompson · Bernoulli bandits, Beta(5,5), δ=0.1 | average number of particles: 6.3 -> 353.4 | Hurt |
Closest papers by meaning
- Thompson Sampling with Less Exploration is Fast and Optimal (ICML 2023 · similarity 0.90)
- On Approximate Thompson Sampling with Langevin Algorithms (ICML 2020 · similarity 0.90)
- The Choice of Noninformative Priors for Thompson Sampling in Multiparameter Bandit Models (AAAI 2024 · similarity 0.90)
- Feel-Good Thompson Sampling for Contextual Bandits: a Markov Chain Monte Carlo Showdown (NeurIPS 2025 · similarity 0.89)
- Fixed Confidence Best Arm Identification in the Bayesian Setting (NeurIPS 2024 · similarity 0.88)
- Balancing Performance and Costs in Best Arm Identification (NeurIPS 2025 · similarity 0.88)
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.