Book Image

Deep Reinforcement Learning with Python - Second Edition

By : Sudharsan Ravichandiran
Book Image

Deep Reinforcement Learning with Python - Second Edition

By: Sudharsan Ravichandiran

Overview of this book

With significant enhancements in the quality and quantity of algorithms in recent years, this second edition of Hands-On Reinforcement Learning with Python has been revamped into an example-rich guide to learning state-of-the-art reinforcement learning (RL) and deep RL algorithms with TensorFlow 2 and the OpenAI Gym toolkit. In addition to exploring RL basics and foundational concepts such as Bellman equation, Markov decision processes, and dynamic programming algorithms, this second edition dives deep into the full spectrum of value-based, policy-based, and actor-critic RL methods. It explores state-of-the-art algorithms such as DQN, TRPO, PPO and ACKTR, DDPG, TD3, and SAC in depth, demystifying the underlying math and demonstrating implementations through simple code examples. The book has several new chapters dedicated to new RL techniques, including distributional RL, imitation learning, inverse RL, and meta RL. You will learn to leverage stable baselines, an improvement of OpenAI’s baseline library, to effortlessly implement popular RL algorithms. The book concludes with an overview of promising approaches such as meta-learning and imagination augmented agents in research. By the end, you will become skilled in effectively employing RL and deep RL in your real-world projects.
Table of Contents (22 chapters)
18
Other Books You May Enjoy
19
Index

Summary

We started the chapter by learning what deep Q networks are and how they are used to approximate the Q value. We learned that in a DQN, we use a buffer called the replay buffer to store the agent's experience. Then, we randomly sample a minibatch of experience from the replay buffer and train the network by minimizing the MSE. Moving on, we looked at the algorithm of DQN in more detail, and then we learned how to implement DQN to play Atari games.

Following this, we learned that the DQN overestimates the target value due to the max operator. So, we used double DQN, where we have two Q functions in our target value computation. One Q function parameterized by the main network parameter is used for action selection, and the other Q function parameterized by the target network parameter is used for Q value computation.

Going ahead, we learned about the DQN with prioritized experience replay, where the transition is prioritized based on the TD error. We explored...