Dr. Yuval Tassa, Google Deepmind: "Machine Learning for Motor Control"
Machine Learning for Motor Control
Dr. Yuval Tassa
Dr. Yuval Tassa from Google Deepmind will give a talk on his exciting research that combines machine learning with advanced physics simulation. Tassa is the first author of the seminal 2012 paper "Synthesis and stabilization of complex behaviors through online trajectory optimization", co-author of Deepmind's highly influential paper "Continuous control with deep reinforcement learning", the lead developer of Deepmind Control Suite and also one of the developers of the MuJoCo physics engine that is widely used by robotics and machine learning researchers.
Talk abstract: Following recent achievements in the application of Deep Reinforcement Learning to rich discrete problem domains like Atari and Go, similar techniques are now being applied to solve problems in continuous motor control. I'll provide a brief overview of successful methods and results from DeepMind and elsewhere, and outline some interesting open questions.
See the Facebook event.