Hybrid Memetic Algorithms And Their Change Detection Applications

Evolutionary computation is a kind of adaptive artificial intelligence technology which is simulating evolution and mechanism to solving problem. Its central idea comes from the basic concept:from simple to complex, from low to high biological evolution process. It is a kind of natural, parallel and robust optimization process. At present, Memetic computation (Cultural Evolutionary Computation) is the higher level evolutionary process of evolutionary computation, which has attracted the interest and concern of national researchers.In this paper, a hybrid memetic algorithm is proposed, and it can be used for numerical optimization, kernel clustering and SAR image change detection. The main contributions can be listed as follows:1) A memetic algorithm with double mutation operators is proposed, termed as MADM. In this paper, the algorithm combines two meta-learning systems to improve the ability of global and local exploration. The double mutation operators in our algorithms guide the local learning operator to search the global optimum; meanwhile the main aim is to use the favorable information of each individual to reinforce the exploitation with the help of two meta-learning systems. Finally, we analysis the performance of our algorithm and comparing algorithm LCSA、DELG and CMA-ES through experiments.2) This paper describes a kernel k-means algorithm with multiple neighborhoods memetic algorithm for clustering complex, unlabeled, and linearly non-separable datasets. It termed as KCMA. The kernel function can transform nonlinear data into a high dimensional feature space, and increases the probability of the linear separability of the patterns within the transformed space and simplifies the associated data structure. According to the distribution of the different datasets, three different local searching operators are designed, and they further enhance the ability of global exploration and overcome premature convergence effectively. Finally, to evaluate this method, the performance of the proposed method has been extensively compared with some classical clustering algorithms over a test suit of several interesting data sets, including artificial datasets and UCI datasets.3) For SAR images change detection problem, based on memetic kernel k-means algorithm for SAR images change detection is proposed. SAR image change detection problem as a combinatorial optimization problem can solved by this method, and the method can obtain the final change detection results by obtaining the optimal threshold value. Finally, in order to analyze the performance of the algorithm, the proposed method has been extensively compared with some clustering algorithms.