A clustering-based differential evolution for global optimization

Clustering-Based Differential Evolution with Composite Trial Vector Generation Strategies and Control Parameters

“…In order to more effectively utilize the population of the search space, clustering operation in C-CODE algorithm used periodically [5] execution, this approach is similar to the method [9]. The algorithm CODE also needs time to explore the search space and organize population clusters.…”

“…Subsequently, in order to further accelerate the convergence rate and balance the exploration and exploitation of DE. Cai et al [5] proposed one-step k-means algorithm to improve the performance of DE. And it is described as follows:…”

“…Max-Generation: maximum generation of population The strategy candidate pool : "rand/1/bin";"rand/2/bin"; "current-to-rand/1". The parameter candidate pool :F=1.0,Cr=0.1;F=0.1,Cr=0.9;F=0.8,Cr=0.2; 1) Generate the initial population P randomly 2) Evaluate the objective function values for each individual in P 3) Initialize the generation counter G=1; 4) while G<Max-Generation do 5) Use CODE to update the population(see line 6-11 in Fig 1) …”

“…In the basic of understanding which trial vector generation strategy and control parameters applicable to the global search, A. E. Eiben, who proposed an adaptive control parameters method [3], Wang Yong et al proposed CODE algorithm [4], which made random combination of several effective strategies and some suitable control parameter settings to improve the performance of DE. In addition, for more efficient use of the population information, Cai et al proposed the CDE algorithm [5] to use one-step k-means as a multi crossover operator.…”

Vinayaka : A Semi-Supervised Projected Clustering Method Using Differential Evolution

“…Engineering problems like aerodynamic design, mechanical design optimization, design of digital filters and multiprocessor synthesis have been solved by DE [2]. Differential Evolution was applied for solving clustering problems [3].…”