mlsanityData to 7 Oct 2026

Thompson Sampling with a Mixture Prior

arXiv 2106.05608 ↗First posted 10 Jun 2021cs.LG cs.AI stat.MLcited by 15

Not on the map. Why?

Abstract (arXiv)

We study Thompson sampling (TS) in online decision making, where the uncertain environment is sampled from a mixture distribution. This is relevant in multi-task learning, where a learning agent faces different classes of problems. We incorporate this structure in a natural way by initializing TS with a mixture prior, and call the resulting algorithm MixTS. To analyze MixTS, we develop a novel and general proof technique for analyzing the concentration of mixture distributions. We use it to prove Bayes regret bounds for MixTS in both linear bandits and finite-horizon reinforcement learning. Our bounds capture the structure of the prior, depend on the number of mixture components and their widths. We also demonstrate the empirical effectiveness of MixTS in synthetic and real-world experiments.

Key facts. Cited by 15 papers in the map (data through 7 Oct 2026). Builds on 18 of its 42 references that are in the map. First posted on arXiv 10 Jun 2021.

Builds on

18 of 42 references.

References · 18

  1. 2014ImageNet Large Scale Visual Recognition Challengecited by 8,441
  2. 2019EfficientNet: Rethinking Model Scaling for Convolutional Neural Networkscited by 6,415
  3. 2020Sharpness-Aware Minimization for Efficiently Improving Generalizationcited by 1,404
  4. 2019Self-training with Noisy Student improves ImageNet classificationcited by 1,304
  5. 2017Revisiting Unreasonable Effectiveness of Data in Deep Learning Eracited by 916
  6. 2013Learning to Optimize Via Posterior Samplingcited by 253
  7. 2013(More) Efficient Reinforcement Learning via Posterior Samplingcited by 233
  8. 2020Model-Based Reinforcement Learning with Value-Targeted Regressioncited by 233
  9. 2014An Information-Theoretic Analysis of Thompson Samplingcited by 178
  10. 2016Corralling a Band of Bandit Algorithmscited by 99
  11. 2019Model selection for contextual banditscited by 60
  12. 2019Information-Theoretic Confidence Bounds for Reinforcement Learningcited by 51
6 more
  1. 2013Thompson Sampling for Complex Bandit Problemscited by 43
  2. 2020Latent Bandits Revisitedcited by 43
  3. 2020Corralling Stochastic Bandit Algorithmscited by 23
  4. 2016Latent Contextual Bandits and their Application to Personalized Recommendations for New Userscited by 18
  5. 2017Variational inference for the multi-armed contextual banditcited by 15
  6. 2015Bayesian Policy Reusecited by 8

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 1; 2022: 3; 2023: 3; 2024: 2; 2025: 3; 2026: 3.

15 papers.

Sort:

By area:

  1. 2021Hierarchical Bayesian Banditscited by 33
  2. 2022Surrogate modeling for Bayesian optimization beyond a single Gaussian processcited by 22
  3. 2022Bayesian Fixed-Budget Best-Arm Identificationcited by 12
  4. 2023Thompson Sampling with Diffusion Generative Priorcited by 7
  5. 2023Exploiting Causal Graph Priors with Posterior Sampling for Reinforcement Learningcited by 5
  6. 2025Contextual Thompson Sampling via Generation of Missing Datacited by 5
  7. 2022Multi-Environment Meta-Learning in Stochastic Linear Banditscited by 4
  8. 2024Solving Multi-Model MDPs by Coordinate Ascent and Dynamic Programmingcited by 2
  9. 2024Online Posterior Sampling with a Diffusion Priorcited by 2
  10. 2025Adaptive Prior Selection in Gaussian Process Bandits with Thompson Samplingcited by 1
  11. 2023Generalized Regret Analysis of Thompson Sampling using Fractional Posteriorscited by 0
  12. 2025Leveraging priors on distribution functions for multi-arm banditscited by 0
3 more
  1. 2026Bayesian Online Model Selectioncited by 0
  2. 2026Posterior Tempering Explains Variance Inflation in Linear and Generalized Linear Thompson Samplingcited by 0
  3. 2026Minimax-Optimality of Posterior Sampling for Reinforcement Learningcited by 0