The automatic control of a dynamic system—for example, a motor, an industrial plant, or a biological function, such as a heartbeat—aims to modify the behavior of the system that is to be controlled, or its outputs, through the manipulation of appropriate quantities of the inputs into the system.
Neural networks are exceptionally effective at generating results that meet the criteria for highly structured data. We could then represent our Q-function with a neural network, which takes the status and action as inputs and then outputs (gives) the corresponding Q-value. Deep reinforcement learning methods use deep neural networks to approximate the reinforcement learning components of the value function, policy, and model. In this chapter, we will learn how to use deep reinforcement learning methods for balancing a...