Speaker: Andr¨¦ T. Martins
Nonextensive entropic kernels
Positive definite kernels on probability measures have been recently applied in structured data classification problems. Some of these kernels are related to classic information theoretic quantities, such as mutual information and the Jensen©\Shannon divergence. Meanwhile, driven by recent advances in Tsallis statistics, nonextensive generalizations of Shannon's information theory have been proposed. This paper bridges these two trends. We introduce the Jensen©\Tsallis q©\difference, a generalization of the Jensen©\Shannon divergence. We then
define a new family of nonextensive mutual information kernels, which allow weights to be assigned to their arguments, and which includes the Boolean, Jensen©\Shannon, and linear kernels as particular cases. We illustrate the performance of these kernels on text categorization tasks.
Andr¨¦ Martins is a PhD student at IST/UTL and SCS/CMU. He is enrolled in the dual CMU©\Portugal PhD program in Language Technologies, under supervision of Mario Figueiredo, Pedro Aguiar (from IST/UTL), Noah Smith and Eric Xing (from CS/CMU). His area of research is "Kernel methods for Natural Language Processing".
Tuesday, July 1st 2008, 14:00 pm
Torre Norte, EA3, Instituto Superior T¨¦cnico