March 23, 2012, Geoffrey Gordon, Carnegie Mellon University
Friday, March 23, 2012 at 3:00 PM
8 St. Mary’s Street, Room 203
Refreshments served at 2:45.
Geoffrey Gordon
Carnegie Mellon University
Spectral Learning Algorithms for Dynamical Systems
If we hope to build an intelligent agent, we have to solve (at least!) the following problem: by watching an incoming stream of sensor data, hypothesize a model of the external world that explains the data. An appealing representation for such a model is a dynamical system—a recursive rule for updating a state, a concise summary of past experience that we can use to predict future observations. Unfortunately, to discover the right dynamical system, we must solve difficult temporal and structural credit assignment problems; often, the result is a search space with a host of bad local optima, and disappointing performance in practice. However, recent work on spectral learning algorithms may provide an end run around these problems: these new spectral algorithms are computationally efficient, simple to implement, statistically consistent, and have no local optima. Perhaps even more interesting, they unify and generalize a number of other state-of-the-art learning methods, including Tomasi-Kanade structure from motion, subspace identification for Kalman filters and hidden Markov models, a novel approach to range-only SLAM, and Laplacian Eigenmaps for manifold learning.
Dr. Gordon is an Associate Research Professor in the Department of Machine Learning at Carnegie Mellon University, and co-director of the Department’s Ph. D. program. He works on multi-robot systems, statistical machine learning, game theory, and planning in probabilistic, adversarial, and general-sum domains. His previous appointments include Visiting Professor at the Stanford Computer Science Department and Principal Scientist at Burning Glass Technologies in San Diego. Dr. Gordon received his B.A. in Computer Science from Cornell University in 1991, and his Ph.D. in Computer Science from Carnegie Mellon University in 1999.
Hosting Professor: Venkatesh Saligrama
Student Host: Justin Foster