An Artificial Immune System with Feedback Mechanisms for Effective Handling of Population Size

SUMMARY This paper represents a feedback artificial immune system(FAIS). Inspired by the feedback mechanisms in the biological immunesystem, the proposed algorithm effectively manipulates the population sizeby increasing and decreasing B cells according to the diversity of the cur-rent population. Two kinds of assessments are used to evaluate the diversityaiming to capture the characteristics of the problem on hand. Furthermore,the processing of adding and declining the number of population is de-signed. The validity of the proposed algorithm is tested for several trav-eling salesman benchmark problems. Simulation results demonstrate theefficiency of the proposed algorithm when compared with the traditionalgenetic algorithm and an improved clonal selection algorithm. key words: artificial immune system, feedback mechanism, population,diversity 1. Introduction Artificial Immune Systems (AIS) are adaptive systems in-spired by theoretical immunology and observed immunefunctions, principles and models [1]. They typically exploitthe immune system’s characteristics of learning and mem-ory in order to develop adaptive systems capable of per-forming a wide range of tasks in various engineering appli-cations, such as anomaly detection [2], pattern recognition[3], data mining [4], traveling salesman problems [5], jobshop scheduling problems [6] and numerical optimizationproblems [7].Within the field of AIS, there are many different typesof algorithm, and research to date has focussed primarily onthe theories of immune networks, clonal selection and neg-ative selection. A large part of the existing work about AIShas been based on the clonal selection theory. Some of thefirst work, known as CLONALG, in applying clonal selec-tion theory was undertaken in the area of optimization prob-lems [5]. Later works concentrated on its improvements interms of the expanded search space [8], the receptor editingoperator [9], the incorporated chaos [10] and so on. Never-theless, little attention has been paid to the population size.Since robustness and computation cost are affected by thepopulation size, it plays a very important role in designingalgorithms. Large population size increases computationalefforts and may make slow convergence, while small size