Generalised Entropy MDPs and Minimax Regret

2014/12/10 by Emmanouil Androulakis, Emmanouil G. Androulakis, Androulakis, Emmanouil G. +2 · 1 citation
Computer Science · Decision Sciences · Mathematics · Physics and Astronomy · #Advanced Bandit Algorithms Research #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Model Reduction and Neural Networks #Neural Networks and Applications #cs.LG #stat.ML

paper · pdf · doi:10.48550/arxiv.1412.3276

7 pages, NIPS workshop "From bad models to good policies"

arxiv created 2014/12/10 · openalex publication_date 2014/12/10 · arxiv updated 2014/12/11 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Bayesian methods suffer from the problem of how to specify prior beliefs. One interesting idea is to consider worst-case priors. This requires solving a stochastic zero-sum game. In this paper, we extend well-known results from bandit theory in order to discover minimax-Bayes policies and discuss when they are practical.

Cited by

• An Asymptotically Optimal Policy for Uniform Bandits of Unknown Support

Related

• Bayesimax Theory: Selecting Priors by Minimizing Total Information
• An Information-Theoretic Approach to Minimax Regret in Partial Monitoring
• Information-Theoretic Minimax Regret Bounds for Reinforcement Learning based on Duality
• An Information-Theoretic Analysis of Bayesian Reinforcement Learning
• An Information-Theoretic Analysis for Thompson Sampling with Many Actions