Back to Search Start Over

An immune network with TH cell function and its applications to pattern recognition.

Authors :
Tamura, Hiroki
Ukon, Satoshi
Tang, Zheng
Ishii, Masahiro
Source :
Electronics & Communications in Japan, Part 3: Fundamental Electronic Science; Oct2005, Vol. 88 Issue 10, p12-22, 11p
Publication Year :
2005

Abstract

The immune function of living bodies has gained attention in engineering applications in a variety of fields. The authors proposed an immune network based on the immune response and researched its application to pattern recognition. The T<subscript>H</subscript> cell (helper T cell) that recognizes antigens in the immune response possesses diversity in order to fight various antigens and the function called interaction that communicates mutual information for correctly recognizing the antigens. The T<subscript>H</subscript> cell function successfully performs self-defense against unknown antigens. This paper proposes an immune network with the T<subscript>H</subscript> cell function added to an immune network we previously proposed. Specifically, we modeled the interaction and diversity functions of the T<subscript>H</subscript> cell in the matching process between input patterns and memory patterns. The proposed immune network is applied to a pattern recognition system of the alphabet and compared to a conventional method by numerical simulations. The proposed immune network exhibits lower incorrect recognition and rejections, and further improvement in the recognition rate over a conventional immune network by employing the process modeling the T<subscript>H</subscript> cell function. © 2005 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 88(10): 12–22, 2005; Published online in Wiley InterScience (<URL>www.interscience. wiley.com</URL>). DOI 10.1002/ecjc.20162 [ABSTRACT FROM AUTHOR]

Details

Language :
English
ISSN :
10420967
Volume :
88
Issue :
10
Database :
Complementary Index
Journal :
Electronics & Communications in Japan, Part 3: Fundamental Electronic Science
Publication Type :
Academic Journal
Accession number :
16905194
Full Text :
https://doi.org/10.1002/ecjc.20162