A Look Inside The Artificial Immune Algorithm Inspired by Clonal Selection Principle

Abstract

Artificial Immune Systems inspired by clonal selection principle (called clonal selection algorithms) have already been successfully applied to pattern recognition tasks. In this paper we present our implementation of one of them, called CLONCLAS, and discuss its behavior in application to recognition of a set of binary patterns. The algorithm performs process of learning based on a set of training data including patterns which belong to ten previously unknown classes and finally generates a group of classifiers which are able to assign the testing input patterns to appropriate classes. Our experiments were performed for a set of commonly known similarity measures of binary strings to select the most efficient of them. We also observed a phenomenon of transformation of memory contents in subsequent phases of iterated process of the system learning.