摘要
强化学习是机器学习领域内的研究热点,主要用来实现决策优化。首先介绍了强化学习的基本原理和经典算法,包括基于值函数的强化学习算法和基于直接策略搜索的强化学习算法;然后针对强化学习目前受关注较多的3个方向:深度强化学习、元强化学习和逆向强化学习分别进行阐述。最后总结了强化学习目前已有的应用和未来可能发展的方向。
Reinforcement learning is a research hotspot in the field of machine learning. It aims to solve problems of decisionor optimization. This paper systematically introduces basic principles and classical reinforcement learning algorithms, including value function based reinforcement learning algorithms and direct policy search based reinforcement learning. Then threedirections including deep reinforcement learning, meta reinforcement learning, inverse reinforcement learning are described.Finally, existing application and development directions of reinforcement learning are summarized.
作者
马骋乾
谢伟
孙伟杰
MA Cheng-qian;XIE Wei;SUN Wei-jie(National University of Defense Technology,Wuhan 430019,China)
出处
《指挥控制与仿真》
2018年第6期68-72,共5页
Command Control & Simulation
关键词
强化学习
深度强化学习
元强化学习
逆向强化学习
决策优化
reinforcement learning
deep reinforcement learning
meta reinforcement learning
inverse reinforcement learning
decision and optimization
