Abstract:

To solve the problems of traditional evolutionary algorithms in computational efficiency, a co-evolution model integrated with an immune mechanism was proposed by referring the idea of co-evolution algorithm. The model maintained the diversity of a population through the respective evolution of multiple sub-populations. During the evolution in each iteration, each sub-population selected the elite antibodies individually and carried out the immune memory operation. Then every sub-population independently mutated with a variety of the algorithm. If the mutation reduced the fitness of the antibody, the antibody was guided by the elite ones. Group collaboration included randomized crossover of a number of individual between sub-populations and large-scale migration among sub-populations. Final the immune metabolism operation removed the weak antibodies in the population. The above operations were repeated until the algorithm reached the established goals or intended loop iterations. Simulation experiments with 13 benchmark functions showed that the optimal solution or satisfactory solution of the model obtained from the search was better than traditional evolutionary algorithms, and its optimization efficiency was also greatly improved.

Key words: co-evolution, immune, evolutionary algorithm, migration, computational efficiency