Hands-on Reinforcement Learning with Ray RLlib
Hands-on Reinforcement Learning with Ray RLlib


Ray RLlib implements a wide variety of reinforcement learning algorithms and it provides the tools for adding your own. It integrates with popular frameworks like OpenAI Gym, TensorFlow, and PyTorch. It provides concise abstractions for defining the algorithm and tools you want to use, and specifying the cluster resources available. It is extensible for new algorithms, agents, and environments. Ray does the work to leverage the resources, providing state-of-the-art performance.

Session Outline
This hands-on tutorial teaches you RLlib with the following lessons:
Bipedal Walker: A popular OpenAI Gym environment, used to introduce RLlib concepts.
Optimizing Market Investments with Multi-Armed Bandits. Using bandits with RLlib and different exploration-exploitation strategies.
RL for Recommender Systems: A more advanced example that explores how to customize RLlib for special needs.

Background Knowledge
Basic understanding of Reinforcement Learning Concepts. One may attend the Reinforcement Learning workshop by Leonardo De Marchi before this session.


Dean Wampler is an expert in data engineering for scalable streaming data systems and applications of machine learning and artificial intelligence (ML/AI). He is a Principal Software Engineer at Domino Data Lab. Previously he worked at Anyscale and Lightbend, where he worked on scalable ML with Ray and distributed streaming data systems with Apache Spark, Apache Kafka, Kubernetes, and other tools. Dean is the author of "Programming Scala", "What Is Ray?", "Fast Data Architectures for Streaming Applications", "Functional Programming for Java Developers", and the coauthor of "Programming Hive", all from O'Reilly. He is a contributor to several open source projects, a frequent conference speaker. He also co-organizes several conferences around the world and several user groups in Chicago. Dean has a Ph.D. in Physics from the University of Washington. Find Dean on Twitter: @deanwampler.