Linear Bayes policy for learning in contextual-bandits

作者：

Highlights：

• A new selection rule for Contextual Bandits and single-step Reinforcement Learning.

• The use of empirical approximations to the Bayes’rule is an effective approach.

• This technique arrived second at new Challenges for Exploration & Exploitation 3.

摘要

•A new selection rule for Contextual Bandits and single-step Reinforcement Learning.•The use of empirical approximations to the Bayes’rule is an effective approach.•This technique arrived second at new Challenges for Exploration & Exploitation 3.

论文关键词：Contextual bandits,Online advertising,Recommender systems,One-to-one Marketing,Empirical Bayes

论文评审过程：Available online 18 July 2013.

论文官网地址：https://doi.org/10.1016/j.eswa.2013.07.041