Differential evolution with dynamic combination based mutation operator and two-level parameter adaptation strategy. (15th April 2022)

• Record Type: Journal Article
• Authors: Deng, Libao; Li, Chunlei; Lan, Yanfei; Sun, Gaoji; Shang, Changjing
• Is Part Of: Expert systems with applications. Volume 192(2022)
• Publication Date: 2022-04-15
• Subjects: Differential evolution -- Mutation operator -- Two-level parameter -- Numerical optimization; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2021.116298
• Languages: English
• ISSNs: 0957-4174

Abstract

Abstract: Differential evolution (DE) is a simple yet effective algorithm for numerical optimization, and its performance significantly depends on mutation operator and control parameters. Therefore, designing appropriate mutation operator and parameter regulation strategy is an important and necessary task. To improve the performance of DE algorithm, we propose a novel DE variant called DCDE based on a dynamic combination based mutation operator and a two-level parameter regulation strategy. More specifically, the newly proposed mutation operator contains a dynamic base vector that consists of two individuals, one is the current optimal individual while the other, called elite individual, is the best one among three randomly selected individuals, and they are dynamically combined by a weight parameter associated with the evolution process and the ranking status of the elite individual in the current population. Moreover, the scale factor and crossover rate in DCDE depend on the combined effect of a population-level parameter and one individual-level parameter, respectively. Both mutation operator and control parameters in DCDE are designed to achieve an appropriate balance between global exploration ability and local exploitation ability. To evaluate the performance of DCDE, comparison experiments are conducted with five state-of-the-art DE variants and three non-DE algorithms on solving 29 functions in IEEE CEC 2017 benchmark suite. The comparison results indicate that the proposed DCDE is significantly better than, or at least comparable to the adopted competitors. Highlights: Combining elite individual and optimal individual to propose a new mutation operator. Combining a macro parameter and a micro parameter to adjust the core parameters. Relationship between operator and parameter is considered when designing algorithm. Comparison results show that the proposed DCDE algorithm has superior performance.