Probabilistic kernel regression models

Authors:

Tommi S. Jaakkola
Department of Computer Science and Electrical Engineering
Massachusetts Institute of Technology
Cambridge, MA 02139
E-mail: tommi@ai.mit.edu

David Haussler
Department of Computer Science
University of California
Santa Cruz, CA 95064
E-mail: haussler@cse.ucsc.edu

Abstract:

We introduce a class of flexible conditional probability models and techniques for classification/regression problems. Many existing methods such as generalized linear models and support vector machines are subsumed under this class. The flexibility of this class of techniques comes from the use of kernel functions as in support vector machines, and the generality from dual formulations of standard regression models.

Keywords:

Gaussian processes, support vector machines, logistic regression, kernel methods

Availability:

(gzipped) PostScript

Other information:

The work was done while T. Jaakkola was at UC Santa Cruz.