ABSTRACT
This chapter introduces the realm of reinforcement learning, provides couple of widespread reinforcement learning examples, and focuses on the development of reinforcement learning algorithms. Reinforcement learning is different from the traditional machine learning methods. The emergence of reinforcement learning dates back to 1850s when Alexander Bain addressed the phenomenon of learning by experiment. A combination of search and memory poses an essential element to reinforcement learning. The two terms have been replaced by exploration and exploitation in modern reinforcement learning. In the early ages of reinforcement learning, several researchers explored trial-and-error learning as an engineering principle. To evaluate a reinforcement learning algorithm in simulations, one has to create an environment and the agent-environment interface. A nice survey on deep reinforcement learning covers a variety of deep reinforcement learning algorithms including the deep Q-network, trust region policy optimization, and asynchronous advantage actor-critic.
