Python Deep Learning Projects

Chapter 14 – Develop an Autonomous Agent with Deep Reinforcement Learning

In this project, we built a deep reinforcement learning model to successfully play the game of CartPole-v1, from OpenAI Gym. Demonstrating mastery here first, we could then extend it to other complex games, such as those by Atari.

We learned how to interact with the Gym toolkit, Q-learning, and SARSA learning; how to code the reinforcement learning model and define hyperparameters; and how to build the training loop and test the model. We found that our SARSA model performed quite a bit better than the Q-learning model. Further training and tuning of hyperparameters, and our own capture of reinforcement units (better scores by our models), should shape our behavior ...