The Advance of Reinforcement Learning and Deep Reinforcement Learning
The development of reinforcement learning is introduced, including classic reinforcement learning methods and deep reinforcementlearning methods, and the challenges faced by reinforcement learning are discussed.
Abstract
Reinforcement learning is one of the leading research fields of artificial intelligence. Unlike other machine learning methods, reinforcement learning is learning from the environment to action mappings. Thus, the chosen action could maximize the accumulated reward value from the environment and develop an optimal strategy via trial-and-error. In recent years, the achievements of deep reinforcement learning represented by AlphaGo have attracted wide attention from researchers. This paper first introduces the development of reinforcement learning, including classic reinforcement learning methods and deep reinforcement learning methods. Then, this paper discusses the advanced reinforcement learning work at present, including distributed deep reinforcement learning algorithms, deep reinforcement learning methods based on fuzzy theory, Large-Scale Study of Curiosity-Driven Learning, and so on. Finally, this essay discusses the challenges faced by reinforcement learning.