The final project in our book is unlike anything we've done so far, and deserves its own treatment. Robotic process automation and optimization, and autonomous agents, such as drones and vehicles, require our deep learning models to learn from environmental cues in a reinforcement learning paradigm. Unlike previous projects, where we've been primarily focused on solving supervised learning problems, in this chapter, we learned to build and train a deep reinforcement learning model capable of playing games.
We employed a deep Q-learning and deep state-action-reward-state-action (SARSA) learning model. Unlike programming simple models by defining heuristics, deep learning models mapping A-B in a supervised learning environment, or determining decision boundaries in cluster analysis in unsupervised learning...