Reinforcement learning (RL) is emerging as a promising approach to intelligently interact with continuously changing physical or virtual environments. Advances in RL research have already shown remarkable results, such as Google’s AlphaGo beating the Go world champion, and are finding their way into self-driving cars, unmanned aerial vehicles, and surgical robotics. Not surprisingly, many see RL growing rapidly into a potentially dominant area in ML over the next decade. However, the applications of RL pose a new set of requirements, the combination of which creates a challenge for existing distributed execution frameworks: computation with millisecond latency at high throughput, adaptive construction of arbitrary task graphs, and execution of heterogeneous kernels over diverse sets of resources.
Ion Stoica, Robert Nishihara, Richard Liaw, Eric Liang, and Philipp Moritz lead a deep dive into Ray, a new distributed execution framework for reinforcement learning applications developed by machine learning and systems researchers at UC Berkeley’s RISELab, walking you through Ray’s API and system architecture and sharing application examples, including several state-of-the art RL algorithms.
Richard Liaw is a PhD student in the Berkeley Artificial Intelligence Research (BAIR) Lab and RISELab at UC Berkeley working with Joseph Gonzalez, Ion Stoica, and Ken Goldberg. He’s worked on a variety of different areas, ranging from robotics to reinforcement learning to distributed systems. He’s working on Ray, a distributed execution engine for AI applications; RLlib, a scalable reinforcement learning library; and Tune, a distributed framework for model training.
Eric Liang is a second-year PhD student working in the University of California, Berkeley, RISELab with Ion Stoica. He works on frameworks and applications for machine learning and reinforcement learning. Previously, he spent several years working on systems in industry at Databricks and Google.
©2018, O’Reilly UK Ltd • (800) 889-8969 or (707) 827-7019 • Monday-Friday 7:30am-5pm PT • All trademarks and registered trademarks appearing on oreilly.com are the property of their respective owners. • email@example.com