Decision Making and Manifolds in Intelligent Systems

EECS Colloquium

Wednesday, September 14, 2016

306 Soda Hall (HP Auditorium)
4:00 – 5:00 pm

Daniel Lee

Professor, School of Engineering and Applied Science
Director, GRASP (General Robotics Automation, Sensing, Perception) Lab
UPS Foundation Chair
University of Pennsylvania

Abstract

Current AI systems for perception and action incorporate a number of techniques: Bayesian state estimation, probabilistic mapping, trajectory planning, and feedback control.  I will describe and demonstrate some of these methods on several autonomous systems including wheeled, legged, and flying robots.  In order to model data variability due to pose, illumination, and background changes, low-dimensional manifold representations have long been used in machine learning.  But how well can such manifolds be processed by neural networks?  I will show how traditional notions of linear separability and VC dimension can be generalized from input points to manifolds.  This analysis provides theoretical predictions for the capacity and generalization ability of invariant classifiers, and better understanding of the performance of deep neural networks.

Biography

Daniel Lee is the UPS Foundation Chair Professor in the School of Engineering and Applied Science at the University of Pennsylvania. He received his B.A. summa cum laude in Physics from Harvard University and his Ph.D. in Condensed Matter Physics from the Massachusetts Institute of Technology in 1995.  Before coming to Penn, he was a researcher at AT&T and Lucent Bell Laboratories in the Theoretical Physics and Biological Computation departments.  He is a Fellow of the IEEE and AAAI and has received the National Science Foundation CAREER award and the University of Pennsylvania Lindback award for distinguished teaching.   He was also a fellow of the Hebrew University Institute of Advanced Studies in Jerusalem, an affiliate of the Korea Advanced Institute of Science and Technology, and organized the US-Japan National Academy of Engineering Frontiers of Engineering symposium.  As director of the GRASP Laboratory and co-director of the CMU-Penn University Transportation Center, his group focuses on understanding general computational principles in biological systems, and on applying that knowledge to build autonomous systems.