Improvement of multi-objective genetic algorithms using differential evolution
П. В. Казаков · Journal of Physics Conference Series · 2021
The following paper describes the manner to improve the efficiency of multi-objective genetic algorithms. It uses the differential evolution with the developed original scheme. One is exploited in the process of creating new individuals. By this scheme in recombination process the searching path is generated into two spaces – decisions and objectives. To increase the probability of creating individuals that dominate their parents various combinations of individuals selected for recombination may be examined. Ultimately it was found that such a strategy should allow to realize a higher replacement of the population dominated individuals. The proposed scheme of differential evolution is used as an extension (modification) for all evolutionary methods of multi-objective optimization. The experiments for the evaluation of the given scheme were conducted on the DTLZ benchmark problems. As a result, in many cases the algorithms which used the proposed scheme of differential evolution, allowed to improve the results (on some indicators) of algorithms SPEA2, NSGA-II and their modifications with the general scheme of differential evolution.