Summary

RL is one of the fundamental branches of machine learning and is currently one of the hottest, if not the hottest, areas of research and development. RL-based AI breakthroughs such as AlphaGo from Google's DeepMind have further increased enthusiasm and interest in the field. This chapter provided an overview of RL and DRL and walked us through a hands-on exercise of building a DQN model using PyTorch.

First, we briefly review the basic concepts of RL. We then explored the different kinds of RL algorithms that have been developed over the years. We took a closer look at one such RL algorithm – the Q-learning algorithm. We then discussed the theory behind Q-learning, including the Bellman equation and the epsilon-greedy-action mechanism. We also explained how Q-learning differs from other RL algorithms, such as policy optimization methods.

Next, we explored a specific type of Q-learning model – the deep Q-learning model. We discussed the key concepts...