The Machine Learning Center at Georgia Tech (ML@GT) has had a wide variety of guest speakers for our Fall Seminar Series. With the semester drawing to a close, we thought we would throw it back all the way to the very first seminar this semester. In fact, this seminar occurred on the first day of classes at Georgia Tech.
Pieter Abeel of the University of California, Berkeley, was kind enough to brave the Georgia summer heat and make a trip to Atlanta to be a part of ML@GT and Georgia Tech’s Institute for Robotics and Intelligent Machines joint seminar.
Abbeel presented his talk, “Deep Learning to Learn” to an auditorium that was bursting with students and faculty.
With plenty of charisma and comedy, Abeel discussed his recent work on meta-learning for action.
His full abstract states, “Reinforcement learning and imitation learning have seen success in many domains, including autonomous helicopter flight, Atari, simulated locomotion, Go, robotic manipulation. However, sample complexity of these methods remains very high. In contrast, humans can pick up new skills far more quickly. To do so, humans might rely on a better learning algorithm or on a better prior (potentially learned from past experience), and likely on both. In this talk, I will describe some recent work on meta-learning for action, where agents learn the imitation/reinforcement learning algorithms and learn the prior. This has enabled acquiring new skills from just a single demonstration or just a few trials. While designed for imitation and RL, our work is more generally applicable and also advanced the state of the art in standard few-shot classification benchmarks such as omniglot and mini-imagenet.”
To see the recording of Abeel’s talk click here.
To see what else is on the schedule for ML@GT’s Fall Seminar Series please visit our website.