Transfer in Reinforcement Learning Via Shared Q-Network

摘要：

　　Transfer learning is one of the hot issues in the field of reinforcement learning.Many reinforcement learning algorithms can only solve a single special task,and their models and algorithms are designed for a single state space,which is not universal.When tasks change,it will cost a lot to change the model and configuration of reinforcement learning algorithm,or even it is not feasible.Transfer learning is born to solve this problem.Through the transfer of learning experience or learning method,the machine learning algorithm can keep the effectiveness and efficiency in the constantly updated tasks and data.

关键词： Transfer Reinforcement Learning Q-network

作者: Zongze Li Zijia Yan Tingting Zhang Yuqiang Qu

作者单位: Department of Software Engineering,Jilin University,Changchun,China

会议类型: 国际会议

会议名称: 2019 4th International Conference on Automatic Control and Mechatronic Engineering (ACME 2019) 2019年第四届自动控制与机电工程国际会议(ACME 2019)

会议地点: 重庆

会议语种:英文

页码: 269-275

在线出版日期: 2019-05-30（万方平台首次上网日期，不代表论文的发表时间）