Reinforcement Learning Pocket Reference

Book description

When it comes to pattern recognition, machine learning and deep learning are excellent--until the underlying data changes. Training ML models to make decisions in a dynamic, ever-changing environment requires reinforcement learning. In this pocket reference, author Matt Kirk shows data scientists, data engineers, and software developers how to apply reinforcement learning to real-world situations.

Despite its long history in academia, reinforcement learning has yet to reach practical business applications. You'll explore how modeling data over time can apply to recommendations, dynamic pricing, medical treatment plans, customer personalization, and traffic flow. This guide includes an easy-to-reference checklist.

You'll explore how to:

  • Build recommendation systems for products or content using bandits
  • Personalize content for customers using contextual bandits
  • Dynamically price ecommerce products using Q-learning and Deep Q-Networks
  • Build a chatbot dialogue engine using policy gradients
  • Apply multiple treatments over time with actor-critic algorithms
  • Optimize traffic flow in a network using Monte Carlo tree search
  • Segment customers based on implied rewards and inverse reinforcement learning

Publisher resources

View/Submit Errata

Product information

  • Title: Reinforcement Learning Pocket Reference
  • Author(s): Matt Kirk
  • Release date: November 2021
  • Publisher(s): O'Reilly Media, Inc.
  • ISBN: 9781098101510