Geometric Aspects of Learning from Labeled and Unlabeled Data

While inference from labeled data is one of the traditional problems of machine learning and statistics, it is only recently that we have developed an  understanding of how unlabeled data may be helpful in various inferential problems. It turns out that many aspects of the connection between labeled and unlabeled data can be interpreted geometrically.

In this talk I will discuss certain geometric invariants, centered around the notions of the Laplace operator and the heat equation, and their role in machine learning. I will also discuss theoretical results on reconstructing these objects from sampled data, as well as some recent applications of these ideas to computing volumes of convex bodies in polynomial time.

Dr. Misha Beylkin

Professor of Computer Science, Ohio State University

Date: March 16, 2007 at 1:00 PM

Location: Engineering Building II, Room 1230

The Department of Electrical and Computer Engineering hosts a regularly scheduled seminar series with preeminent and leading reseachers in the US and the world, to help promote North Carolina as a center of innovation and knowledge and to ensure safeguarding its place of leading research.