Luo Yinsheng1, Li Renhou1, Zhang Weixi2
(1. School of Electronics and Information Engineering, Xi'an Jiaotong University, Xi'an 710049, China; 2. Department ofElectrical and Information Engineering, Jiangsu Teachers University of Technology, Changzhou 213001, China)
Abstract: Based on the evolutionary and learning mechanism of immune cells in germinal
center reactions, a new algorithm for dynamic function optimization was proposed, and the
features of multi-population cells, diversity of cells, and recycling and immune memory in
the germinal center were simulated. The advantages of the algorithm are that the multiple
searching subpopulations are produced by the base and clone populations, the cells’hyper-mutation
are inverse proportional to its affinity. Furthermore, it produces and updates the memory
cells pool, and whether the function is changed or not is tested continuously at each
generation. Using moving peaks function as a testing benchmark and the offline average
error as a performance measure of the algorithm, simulation experiments are carried out,
and the results show that the proposed algorithm can approximate the optimum of the
dynamic function with a smaller average error and variance when the function change
frequency is slow.
Keywords: dynamic function; immune system; germinal center reaction; optimization
algorithm