Idiotypic Network Model for Feature Extraction in Pattern Recognition-Effect of Diffusion of Antibody

An idiotypic network model using antibodies as partial templates was applied to a pattern recognition problem. A novel network model that contains diffusion of antibodies was proposed. The proposed technique was applied to classification of 33 samples of Chinese characters that have a common sub-structure in the character. With partial template matching they were classified into 12 groups. Using the dynamics of the proposed network, the samples were classified into 18 groups. Through this study, the applicability of the idiotypic network with antibody diffusion to pattern recognition was verified.