Immune Inspired Restricted Somatic Hypermutation for Multimodal Optimization

摘要

An improved immune optimization algorithm is proposed to solve the contradiction between global search and local optimization which existed in most traditional optimization algorithms for multimodal function. The key idea lies on that hypermutation operator with restriction is designed for parallel search. By simulating the property of metadynamics in immune system, the algorithm can dynamically adjust the population size. In the view of the population size and individual space, the validity of the mutation operator is analyzed by transition probability. It is proved theoretically that the presented algorithm is convergence. The simulation to 4 benchmark functions verified that the algorithm can obtain the multiple local and global optima simultaneously.