TR available: Task Decomposition and Module Combination

[Bao-Liang Lu]

The following Technical Report is available via anonymous FTP.

FTP-host:ftp.bmc.riken.go.jp

FTP-file:pub/publish/Lu/lu-ieee-tnn-98-2nd-rev.ps.gz

==========================================================================
TITLE: Task Decomposition and Module Combination Based on Class Relations:
       A Modular Neural Network for Pattern Classification 

       BMC Technical Report BMC-TR-98-1

AUTHORS:
       Bao-Liang Lu and Masami Ito
     
ORGANISATIONS:
       Bio-Mimetic Control Research Center,
       The Institute of Physical and Chemical Research (RIKEN)

ABSTRACT:
In this paper, we propose a new method for decomposing 
pattern classification problems based on the class relations 
among training data. By using this method, we can divide a 
$K$-class classification problem into a series of ${K\choose 2}$ 
two-class problems. These two-class problems are to discriminate 
class ${\cal C}_{i}$ from class ${\cal C}_{j}$ for $i=1,\,
\cdots,\, K$ and  $j=i+1$, while the existence of the training 
data belonging to the other $K-2$ classes is ignored. If the 
two-class problem of discriminating class ${\cal C}_{i}$ from class 
${\cal C}_{j}$ is still hard to be learned, we can further break 
down it into a set of two-class subproblems as small as we expect. 
Since each of the two-class problems can be treated as a completely 
separate classification problem with the proposed learning paradigm, 
the two-class problems can be learned by different network modules 
in parallel. We also propose two module combination principles 
which give practical guidelines in integrating individual trained 
modules. After learning of each of the two-class problems with 
a network module, we can easily integrate all of the trained modules 
into a min-max modular (${\rm M}^{3}$) network according to the module 
combination principles and obtain a solution to the original problem.
Consequently, a large-scale and complex $K$-class classification 
problem can be solved effortlessly and efficiently by learning a 
series of smaller and simpler two-class problems in parallel.

(38 pages, 4.8Mk)

Any comments are appreciated.

Bao-Liang Lu
======================================
Bio-Mimetic Control Research Center
The Institute of Physical and Chemical Research (RIKEN)
2271-130, Anagahora, Shimoshidami, Moriyama-ku
Nagoya 463-0003, Japan
Tel: +81-52-736-5870
Fax: +81-52-736-5871
Email: lbl at bmc.riken.go.jp