Reinforcement Learning is Becoming Practical
Estimated Reading Time: 2 minutes
TL;DR
We have a short post this week covering a recent advance in practical reinforcement learning followed by a few selected news items from around the web.
Reinforcement Learning for Better Video Compression
A fascinating recent preprint by DeepMind applies their MuZero system to the problem of designing a better video compression algorithm. As the figure below demonstrates, video compression can be formulated as a sequential decision problem of selecting compression rates based on previous scenes. DeepMind proposes a novel self-competition scheme that can be used as a loss for training a better video compression scheme. The trained compression algorithm sees about 6% improvement across a wide range of YouTube videos. Special thanks to Import AI for featuring this paper and bringing it to our attention.
Reinforcement learning continues to mature as a practical technique. The use of RL in practice still remains the province of experts, but additional improvements in the next few years may make these techniques broadly applicable.
Weekly News Roundup
https://pypistats.org/packages/deepchem: DeepChem was downloaded over 17,000 times last month!
https://spectrum.ieee.org/open-robotics: The open source robotics foundation turns 10 years old
https://physics.aps.org/articles/v15/s38: Creating a semiconductor-superconductor interface
https://www.nextplatform.com/2022/03/23/nvidia-will-be-a-prime-contractor-for-big-ai-supercomputers/: Nvidia is becoming a serious competitor in the supercomputing market.
https://www.wsj.com/articles/uber-reaches-deal-to-list-all-new-york-city-taxis-on-its-app-11648123201: Uber to list taxi drivers on its app in New York City
Feedback and Comments
Please feel free to email me directly (bharath@deepforestsci.com) with your feedback and comments!
About
Deep Into the Forest is a newsletter by Deep Forest Sciences, Inc. We’re a deep tech R&D company building an AI-powered scientific discovery engine. Deep Forest Sciences leads the development of the open source DeepChem ecosystem. Partner with us to apply our foundational AI technologies to hard real-world problems. Get in touch with us at partnerships@deepforestsci.com!
Credits
Author: Bharath Ramsundar, Ph.D.
Editor: Sandya Subramanian, Ph.D.