preprint.

[Ke CHEN]

Dear Connectionists,

The following preprint is available now on line:

http://www.cis.ohio-state.edu/~kchen/jnc98.ps

Best regards,

-kc

----------------------------------------------------
Dr. Ke CHEN
Department of Computer and Information Science
The Ohio State University
583 Dreese Laboratories
2015 Neil Avenue
Columbus, Ohio 43210-1277
U.S.A.

Phone:  1-614-292-4890(O) (with an answering machine)
Fax:    1-614-292-2911
E-Mail: kchen at cis.ohio-state.edu
WWW: http://www.cis.ohio-state.edu/~kchen
------------------------------------------------------





##########################################################################
         A Method of Combining Multiple Probabilistic
   Classifiers through Soft Competition on Different Feature Sets

                Ke Chen{1,2} and Huisheng Chi{1}

 {1} National Lab of Machine Perception and Center for Information Science
                  Peking University, Beijing 100871, China

            {2} Dept of CIS and Center for Cognitive Science
        The Ohio State University, Columbus, OH 43210-1277, USA

      To appear in NEUROCOMPUTING - AN INTERNATIONAL JOURNAL, 1998.


                             ABSTRACT

A novel method is proposed for combining multiple probabilistic
classifiers on different feature sets. In order to achieve the improved
classification performance, a generalized finite mixture model is proposed
as a linear combination scheme and implemented based on radial basis
function networks.  In the linear combination scheme, soft competition on
different feature sets is adopted as an automatic feature rank mechanism
so that different feature sets can be always simultaneously used in an
optimal way to determine linear combination weights. For training the
linear combination scheme, a learning algorithm is developed based on
Expectation-Maximization (EM)  algorithm. The proposed method has been
applied to a typical real world problem, viz. speaker identification, in
which different feature sets often need consideration simultaneously for
robustness. Simulation results show that the proposed method yields good
performance in speaker identification. 

Keywords: Combination of multiple classifiers, soft competition,
different feature sets, Expectation-Maximization (EM) algorithm,
speaker identification
##########################################################################