Complexity analysis of reinforcement learning and its application to robotics

Abstract

Reinforcement learning (RL) is a widely adopted theory in machine learning, which aims to handle the optimal decision of intelligent agent interacting with the stochastic dynamic environment. Its origin may come from the motivation of phycological observations since 1960's [1]. It blooms recently as the emerging of large sample data and powerful computation facility, especially the AlphaGo's beat over the human top Go player in 2016 [2].