Multiscale Paradigm in Genetic Algorithm
Yoon Young Kim, Dong Hoon Jung · 9th AIAA/ISSMO Symposium on Multidisciplinary Analysis and Optimization · 2002
In this paper, we propose a new multiscale paradigm genetic algorithm . The motivation of this method is to make a genetic algorithm applicable for very large -size problem s, which are otherwise difficult to solve by standard single -scale genetic algorithms. In developing the multiscale genetic algorithm, all the key ingredients of genetic algor ithms, such as representation, crossover, and mutation, are expressed in multiscales; these multiscale expressions are also developed here for the first time. To handle binary data in multiscales, we employ the binary wavelet transform using modulo -2 opera tions. Since binary data are expressed in muliscales, the multiscale genetic algorithm can be carried out in multiresolution. In some typical large -size problems, the present multiscale genetic algorithm is shown to be much superior to a standard genetic algorithm.