UCSB COMPUTER SCIENCE DEPARTMENT PRESENTS FACULTY CANDIDATE:
BRIAN KULIS Department of Computer Science University of Texas at Austin
DATE: WEDNESDAY, FEBRUARY 27, 2008
TIME: 3:30 – 4:30 p.m.
PLACE: Computer Science Conference Room, Harold Frank Hall Rm. 1132
Several data analysis problems rely on an appropriate distance function to compare data objects. Examples include visual object classification, text retrieval algorithms for search engines, clustering, human body pose estimation, and many others. Typically, the distance functions used in such applications are either off-the-shelf distance functions such as the Euclidean distance, or are hand-tuned distance functions specific to a particular task. In this talk, I will discuss some of my recent work on learning distance functions automatically. In order to make distance learning practical, the resulting algorithm must possess a number of crucial properties, including the ability to i) scale to data sets of millions of objects or more, ii) work with high-dimensional data such as images and text, and iii) support fast similarity searches. I will describe an approach, based on the Log Determinant divergence, which satisfies each of these properties. In addition to developing the optimization techniques necessary for implementation, I will cover some applications of this work, including fast image search and automated software support, as well as extensions and related work.
Brian Kulis is a PhD student in the computer science department at the University of Texas at Austin. His research focuses on machine learning, data mining, and large-scale optimization. He obtained his BA from Cornell University in computer science and mathematics in 2003. For his research, he has won two best student paper awards at the International Conference on Machine Learning, and is also the recipient of an MCD fellowship from the University of Texas (2003-2007) and an Award of Excellence from the College of Natural Sciences at the University of Texas. In the fall of 2007, he was a research fellow at the Institute for Pure and Applied Mathematics at UCLA.
HOST: CHANDRA KRINTZ