Machine Learning and Big Data

11Reinforcement Learning

Amandeep Singh Bhatia¹*, Mandeep Kaur Saggi², Amit Sundas¹ and Jatinder Ashta¹

¹ Chitkara University Institute of Engineering and Technology, Chitkara University, Punjab, India

² Department of Computer Science & Engineering, Thapar Institute of Engineering & Technology, Patiala, India

Abstract

Reinforcement learning (RL) has gradually become one of the most active research areas in the field of artificial intelligence and machine learning (i.e., agent learns to interact with the environment to achieve reward, robotics, and many more). It is a sub-area of machine learning. Due to its generality, it has been studied widely in many other disciplines such as operations research, control theory, game theory, swarm intelligence, and multi-agent systems. In this chapter, the model-free and model-bases RL algorithms are described. There exist several challenges that need to be addressed. One of challenges that arise in RL is trade-off between exploration and exploitation. The dilemma of exploration-exploitation has been intensively presented.

�Keywords: Machine learning, reinforcement learning, Q-learning algorithm, Monte Carlo method, SARSA learning, R-learning, temporal difference, dyna-Q learning

11.1 Introduction: Reinforcement Learning

It is defined as a computational method for compassionating and automating purposive behavior learning and decision making. It represents to solving sequential and to control a stochastic dynamical system by simulation and ...