Deep Reinforcement Learning with Surrogate Agent-Environment Interface.
作者： Yu Jing , Song Wang
DOI:
关键词:
摘要: In this paper, we propose surrogate agent-environment interface (SAEI) in reinforcement learning. We also state that learning based on probability provides optimal policy of task interface. introduce action and develop the deterministic gradient (PSADPG) algorithm SAEI. This enables continuous control discrete action. The experiments show PSADPG achieves performance DQN certain tasks with stochastic nature initial training stage.
arxiv.org
harvard.edu参考文章(3)
Ronald J. Williams, Simple Statistical Gradient-Following Algorithms for Connectionist Reinforcement Learning Machine Learning. ,vol. 8, pp. 229- 256 ,(1992) , 10.1007/BF00992696
Yuval Tassa, Daan Wierstra, Alexander Pritzel, Tom Erez, Jonathan J. Hunt, Nicolas Heess, David Silver, Timothy P. Lillicrap, Continuous control with deep reinforcement learning arXiv: Learning. ,(2015)
Gabriel Dulac-Arnold, Peter Sunehag, Ben Coppin, Richard Evans, Reinforcement Learning in Large Discrete Action Spaces. arXiv: Artificial Intelligence. ,(2015)