Welcome to the chapter on reinforcement learning. In the previous chapters, we have worked on solving supervised learning problems. In this chapter, we will learn to build and train a deep reinforcement learning model capable of playing games.
Reinforcement learning is often a new paradigm for deep learning engineers and this is why we're using the framework of a game for this training. The business use cases that we should be looking out for are typified by process optimization. Reinforcement learning is great for gaming, but also applicable in use cases ranging from drone control (https://arxiv.org/pdf/1707.05110.pdf) and navigation to optimizing file downloads over mobile networks (http://anrg.usc.edu/www/papers/comsnets_2017.pdf).
We will do this with something called deep Q-learning and deep State-Action-Reward-State...