STATISTIC AND APPLIED PROBABILITY
SEMINAR
DATE and TIME: Nov 23-th, 2004 @1 pm in NS 234
SPEAKER: Refaat M. Mohamed (UofL CVIP Lab )
TITLE: Mean Field
Theory for Density Estimation using Support Vector Machines
ABSTRACT:
Recently, Support Vector Machines (SVM) has proven itself as a
promising algorithm for different applications of the pattern
recognition and computer vision community. In this talk, the SVM as a
regression algorithm is presented. The traditional formulation of the
SVM algorithm which raises a quadratic optimization problem will be
discussed. An algorithm which uses the Mean Field theory to approximate
the learning of the SVM algorithm in such a way to avoid raising the
quadratic programming is presented. Experimental results on synthetic
data as well as real remote sensing data illustrate the performance of
the proposed algorithm.