Review of reinforcement learning research

Abstract

Reinforcement learning is a type of machine learning. An important feature that distinguishes it from other types of learning is that reinforcement learning uses training information to evaluate actions taken. The correct action guides the choice of action. The agent is not told what action to do and what action should not be done. Instead, it tries to discover what action can produce the maximum reward. Therefore, reinforcement learning is a trial and error mechanism that learns through constant trial and error and feedback. Corresponding algorithms include dynamic programming, Monte Carlo methods, Q-Learning, TD-Learning, and Sarsa algorithms.

Topics & Concepts