#### Overview of this book

Deep Reinforcement Learning Hands-On is a comprehensive guide to the very latest DL tools and their limitations. You will evaluate methods including Cross-entropy and policy gradients, before applying them to real-world environments. Take on both the Atari set of virtual games and family favorites such as Connect4. The book provides an introduction to the basics of RL, giving you the know-how to code intelligent learning agents to take on a formidable array of practical tasks. Discover how to implement Q-learning on 'grid world' environments, teach your agent to buy and trade stocks, and find out how natural language models are driving the boom in chatbots.
Table of Contents (23 chapters)
Deep Reinforcement Learning Hands-On
Contributors
Preface
Other Books You May Enjoy
Free Chapter
What is Reinforcement Learning?
OpenAI Gym
Deep Learning with PyTorch
The Cross-Entropy Method
Tabular Learning and the Bellman Equation
Deep Q-Networks
DQN Extensions
Stocks Trading Using RL
Policy Gradients – An Alternative
The Actor-Critic Method
Asynchronous Advantage Actor-Critic
Chatbots Training with RL
Web Navigation
Continuous Action Space
Trust Regions – TRPO, PPO, and ACKTR
Black-Box Optimization in RL
Beyond Model-Free – Imagination
AlphaGo Zero
Index

## Actor-critic

The next step in reducing the variance is making our baseline state-dependent (which, intuitively, is a good idea, as different states could have very different baselines). Indeed, to decide about the suitability of a particular action in some state, we're using the discounted total reward of the action. However, the total reward itself could be represented as a value of the state plus advantage of the action: Q(s, a) = V(s) + A(s, a). We've seen this in Chapter 7, DQN Extensions, when we discussed DQN modifications, particularly dueling DQN.

So, why can't we use V(s) as a baseline? In that case, the scale of our gradient will be just advantage A(s, a), showing how this taken action is better in respect to the average state's value. In fact, we can do this, and it is a very good idea for improving the PG method. The only problem here is: we don't know the value of the V(s) state to subtract it from the discounted total reward Q(s, a). To solve this, let's use another neural network...