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

Actor-critic using Kronecker-factored trust region

ACKTR, as the name suggests, is the actor-critic algorithm based on the Kronecker factorization and trust region.

We know that the actor-critic architecture consists of the actor and critic networks, where the role of the actor is to produce a policy and the role of the critic is to evaluate the policy produced by the actor network. We learned that in the actor network (policy network), we compute gradients and update the parameter of the actor network using gradient ascent:

Instead of updating our actor network parameter using the preceding update rule, we can also update it by computing the natural gradients as:

Where F is called the Fisher information matrix. Thus, the natural gradient is just the product of the inverse of the Fisher matrix and standard gradient:

The use of the natural gradient is that it guarantees a monotonic improvement in the policy. However, updating the actor network (policy...