Reinforcement Learning

Abstract

Reinforcement learning allows computational agents to act intelligently in a complex (and often dynamic) environment in order to achieve a narrowly defined goal. The sheer number of complex ideas involved in reinforcement learning makes it more challenging to grasp than the previously discussed supervised learning frameworks. That is why in this chapter, we introduce reinforcement learning by pulling apart the entire process, and by introducing each concept as needed. By focusing our attention on just one piece of the system at a time, we can gain a fuller understanding of how each component works, and why it is needed. Moreover, by doing so, we also gain some extremely important intuition about how the individual components of reinforcement learning define the strengths and limitations of the process in general.