PAC-Bayes Bounds for Bandit Problems: A Survey and Experimental Comparison

Abstract

PAC-Bayes has recently re-emerged as an effective theory with which one can derive principled learning algorithms with tight performance guarantees. However, applications of PAC-Bayes to bandit problems are relatively rare, which is a great misfortune. Many decision-making problems in healthcare, finance and natural sciences can be modelled as bandit problems. In many of these applications, principled algorithms with strong performance guarantees would be very much appreciated. This survey provides an overview of PAC-Bayes bounds for bandit problems and an experimental comparison of these bounds. On the one hand, we found that PAC-Bayes bounds are a useful tool for designing offline bandit algorithms with performance guarantees. In our experiments, a PAC-Bayesian offline contextual bandit algorithm was able to learn randomised neural network polices with competitive expected reward and non-vacuous performance guarantees. On the other hand, the PAC-Bayesian online bandit algorithms that we tested had loose cumulative regret bounds. We conclude by discussing some topics for future work on PAC-Bayesian bandit algorithms.

32 pages, 8 figures

References in corpus (12)

• OpenML: networked science in machine learning
• On Markov chain Monte Carlo methods for tall data
• Doubly Robust Policy Evaluation and Optimization
• A Note on the PAC Bayesian Theorem
• PAC-Bayes Mini-tutorial: A Continuous Union Bound
• PAC-Bayes Generalisation Bounds for Heavy-Tailed Losses through Supermartingales
• PAC-Bayesian Lifelong Learning For Multi-Armed Bandits
• A unified recipe for deriving (time-uniform) PAC-Bayes bounds
• Learning PAC-Bayes Priors for Probabilistic Neural Networks
• Meta-strategy for Learning Tuning Parameters with Guarantees
• PAC-Bayes Iterated Logarithm Bounds for Martingale Mixtures
• Progress in Self-Certified Neural Networks