Preference Multi-objective Optimization Based On Artificial Immune System

The biological immune system is a highly parallel, distributed, adaptive and self organized system. It also has the ability for learning, recognition, memory and feature extraction. In order to better solve the problems related to engineering applications, artificial immune system is developed by simulating the information processing abilities of biological immune system. And it has been successfully applied in the field of multi-objective optimization. Based on the artificial immune system, this paper discuss the preference multi-objective optimization, the main work in this paper are as follows:In chapter two, an adaptive ranks clone and differential evolution based immune multi-objective optimization algorithm is proposed, which is based on the nondominated neighbor immune algorithm (NNIA). This algorithm uses an adaptive ranks clone and a new differential evolution scheme. By using the new differential evolution scheme the dominated solutions in active antibody can learn the information of the non-dominated ones and the unnecessary function evaluations can be decreased. The experimental study proves that this algorithm can get a set of more precise solutions.In chapter three, preference rank immune memory clone selection algorithm PISA is introduced. We proposed a preference multi-objective optimization based on adaptive ranks clone and differential evolution by introducing the new differential evolution scheme and adaptive ranks clone mentioned in chapter two into PISA. The experimental study proves that this algorithm well in terms of generational distance, spacing and hypervolume. However, when the number of objective as high as ten the number of successful convergence obtained by this algorithm is small.In chapter four, we hybridized the Pareto dominance principle and the light beam search method and proposed a new dominated relation, and adaptively control the threshold parameter, keeping the diversity of the population. And proposed a reference based dominance based preference rank immune memory clone selection algorithm (r-PISA). The experimental study proves that r-PISA can always successful convergence to the preference region even the number of objectives as high as eight.This paper was supported by the National Natural Science Foundation (No.60803098) and by the National Research Foundation for the Doctoral Program of Higher Education of China (No.20070701022). The Provincial Natural Science Foundation of Shaanxi of China (2010JM8030). The Fundamental Research Funds for the Central Universities (NO. K50511020014).