Research On Improved Differential Evolution Algorithm For Solving High-Dimensional Global Optimization Problems

Global Optimization Problems (GOPs) are prevalent in science and engineering, and have important research significance and application value. With the development of research, High-dimensional Global Optimization Problems (HGOPs), which have huge search space and are difficult to solve, are becoming a new hotspot. Instead of owning good performance for solving GOPs, Differential Evolutionary (DE) algorithm need to be greatly improved to achieve better results for solving HGOPs. This paper focuses on improving DE to solve HGOPs, and discusses some related areas such as Evolutionary Algorithm Framework (EAF).Firstly, this paper described the standard implementation, existing differential strategies, the latest developments and current research focuses of DE. After explaining the HGOPs, introduced Co-Evolutionary Algorithm (CEA) which has effective performance for solving HGOPs.Combined DE with Potter co-evolution model, this paper proposed the Cooperative Coevolving Differential Evolution with Adaptive Building-blocks Identification (CCDE-ABI) algorithm. CCDE-ABI uses two populations, one is used to permute the gene sequence of solution individuals to identify the building-blocks, the other is used to search for optimal solutions and evaluate the fitness of the gene sequence individual. Two populations using different evolutionary algorithm have Cooperative Co-evolution (CC) mechanism in searching the optimal solution. Experiments show that CCDE-ABI has excellent global and local search capabilities, and is an effective algorithm for solving HGOPs.Since exist many branches of the evolutionary algorithm, the researchers need facilitate research tool. This paper designed a evolutionary algorithms framework named EA++ which abstracts Evolutionary Algorithm (EA) major components such as population structure, evolution process, evolutionary operators, fitness evaluation, statistics etc. EA++ provides the evolutionary algorithm fundamental infrastructure, and has low-coupling modules, scalable hierarchical structure, efficient underlying implementation and more secure exception handling mechanism. Compared to other evolutionary algorithms frameworks, EA++ expresses efficient, flexible, scalable, configurable, portable characteristics, and has higher application value for researchers.