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)
Other Books You May Enjoy

Twin delayed DDPG

Now, we will look into another interesting actor-critic algorithm, known as TD3. TD3 is an improvement (and basically a successor) to the DDPG algorithm we just covered.

In the previous section, we learned how DDPG uses a deterministic policy to work on the continuous action space. DDPG has several advantages and has been successfully used in a variety of continuous action space environments.

We understood that DDPG is an actor-critic method where an actor is a policy network and it finds the optimal policy, while the critic evaluates the policy produced by the actor by estimating the Q function using a DQN.

One of the problems with DDPG is that the critic overestimates the target Q value. This overestimation causes several issues. We learned that the policy is improved based on the Q value given by the critic, but when the Q value has an approximation error, it causes stability issues to our policy and the policy may converge to local optima.