In this chapter, we introduced concepts of reinforcement learning and how they are different from traditional supervised learning techniques. We described the core ideas behind RL, as well as basic modules such as Q-learning and policy learning that characterize any reinforcement learning technique today. We also presented deep learning-based advances to traditional RL techniques in form of DRL. We illustrated various different network architectures for DRL and discussed their relative merits. Finally, we sketched the core implementation of a few reinforcement learning tasks as applied to some popular computer-based games.
In next chapter, we will look at some of the practical tips and tricks used while implementing deep learning models in real world applications.
