Research Of High-dimensional Numerical Optimization Based On Memetic Algorithm

Numerical optimization problems have widely existed in the field of engineering and scientific research. Evolutionary algorithm, as a kind of new heuristic optimization algorithm, has been used more and more widely in the field of numerical optimization, for its simplicity, high efficiency and strong capability of global search, etc. However, most of the evolutionary algorithms suffer from the problem of "the curse of dimensionality"-their performance deteriorates quickly with the grown of dimensionality. In order to solve this problem, in this paper, the method for solving high-dimensional numerical optimization problems have been studied based on Memetic Algorithms (MAs), and the main work is the improvement of local search operator. The main contents and innovation are as the following:1. In the framework of MA, the adaptive local search depth operator (ALSD) has been designed, which can adjust the local search depth dynamically according to the current search state. The experimental results on LSGO test suite issued in CEC’2012show that ALSD can largely improve the performance of MAs.2. In this paper, Cooperative Coevolution (CC), was combined with local search operator. Through a lot of experimentation, CC was embedded into local search under the framework of MA to study its influence on local search in solving different kinds if high-dimensional optimization problems. Some meaningful results have been obtained.