CMU Campus
Center for                           Nonlinear Analysis
CNA Home People Seminars Publications Workshops and Conferences CNA Working Groups CNA Comments Form Summer Schools Summer Undergraduate Institute PIRE Cooperation Graduate Topics Courses SIAM Chapter Seminar Positions Contact
CNA Seminar/Colloquium/Joint Pitt-CNA Colloquium
James von Brecht
UCLA
Title: Total Variation Clustering

Abstract: A popular class of clustering models rely on the minimization of an energy defined over the set of admissible partitions of a data set. These discrete optimizations usually pose NP-hard problems, however. A natural resolution of this issue involves relaxing the discrete minimization space into a continuous one to obtain an easier minimization procedure. Many current algorithms, such as spectral clustering or non-negative matrix factorization, follow this relaxation approach. Ideas from the imaging science literature have recently motivated a new set of algorithms that use total variation relaxations instead. We will summarize the motivation behind the total variation approach, in terms of both theoretical and algorithmic results, and provide a comparison to other state-of-the-art approaches.

Date: Tuesday, September 17, 2013
Time: 1:30 pm
Location: Wean Hall 7218
Submitted by:  David Kinderlehrer