Hybrid Clonal Selection Algorithm For Multi-modal Function Optimization

Multi-modal function optimization is a typical function optimization in scientific researchand engineering practice. Some optimization methods have been proposed, but it is difficult tofind all global optimal solutions by traditional optimization algorithm，because the multi-modalfunctions have more global optimal solutions and local optimal solutions than the single-modalfunctions. The colonal selection algorithms are probabilistic search algorithm inspired fromnatural selection and elimination mechanism, which is parallel, efficient, robust, universal,concise etc. The algorithms have been widely used in pattern recognition, exception and faultdiagnosis, robot control, network intrusion detection, function optimization and other areas, andhave a more superior performance and efficiency.However, the traditional clonal selection algorithm, which uses the "Copy-Cross-mutation" mechanism and adaptive selection by the proportion, usually leads to quickconvergence to local optimal solutions. This convergence is called “premature convergence”.Meanwhile, the traditional clonal selection algorithm uses binary encoding, which need a largenumber of operations on coding and decoding computation, in addition, its convergence accuracyis very low.How to improve the clonal selection algorithm has become a hot research point. This thesisaims to improve the clonal selection algorithm for multi-modal function optimization by hybridstrategy. The main works are as the following, for the multi-modal optimization problems, theresearch ideas of hybrid clonal selection algorithm is proposed. The thesis propose a hybridintelligent optimization algorithm integrating clonal selection and simulated annealing algorithm,the diversity of solutions and its optimization efficiency are analyzed by functions optimizationexperiments. Another hybrid intelligent optimization algorithm integrating clonal selection andparticle swarm optimization algorithm is proposed. The algorithm’s efficiency and validity areverified by its application in function optimization testing experiments.Experiment results show that the hybrid clonal selection algorithms have advantages inparallel computing, computation efficiency and robustness, and they are easy to solve thosecomplex multi-modal functions optimization problems.