RLCFR: Minimize counterfactual regret by deep reinforcement learning

作者：

Highlights：

• Solving game based on counterfactual regret in a reinforcement learning framework.

• Studying on improving generalization ability of counterfactual regret based method.

• A policy is learned by the agent to select appropriate regret updating method.

• The generalization ability of our method is significantly improved.

摘要

•Solving game based on counterfactual regret in a reinforcement learning framework.•Studying on improving generalization ability of counterfactual regret based method.•A policy is learned by the agent to select appropriate regret updating method.•The generalization ability of our method is significantly improved.

论文关键词：Counterfactual regret minimization,Decision-making,Imperfect information,Reinforcement learning

论文评审过程：Received 16 September 2020, Revised 7 July 2021, Accepted 19 September 2021, Available online 30 September 2021, Version of Record 6 October 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2021.115953