Enhanced Cooperative Coevolution Optimization Algorithm Based On Convergence Speed Controller And Its Application

Many real problems can be abstracted into complex large-scale optimization problems.In general,the performance of optimization algorithm decreases quickly as the dimension of the search space increases.In order to effectively solve large-scale optimization problems,researchers have proposed the cooperative coevolution(CC)optimization algorithm to decompose the large-scale optimization problem into a series of small-scale sub-optimization problems,and then use heuristic optimization algorithm to solve them.However,if the search space of problem is complex,there are many associated variables,even variables are non-separable,the decomposed sub-optimization problem may still be a complex optimization problem.If the performance of heuristic algorithm for solving sub-optimization problems is insufficient,the overall performance of CC optimization algorithm will be reduced.To solve the above problems,the research work of this paper is to design strategy to enhance the performance of the heuristic algorithm for solving sub-optimization problems in CC optimization algorithm,and proposed a CC optimization algorithm framework based on convergence speed controller;Besides,the proposed algorithm framework is applied to solve two real large-scale optimization problems.The main work is as follows:(1)According to the existing research results of convergence speed controller and limited computational resources when solving real optimization problems,the periodic detection mechanism of convergence speed controller is improved.The improved convergence speed controller is applied to the CC optimization algorithm to improve the performance of heuristic algorithm for solving sub-optimization problems.The experiments on the open test function prove that convergence speed controller can effectively enhance the performance of the CC optimization algorithm.(2)Based on the proposed algorithm framework,a cooperative coevolution differential evolution algorithm based on convergence speed controller is designed to solve the large-scale sample optimization problem in the image matting technique.By comparing with the existing optimization algorithms on the open image test set,the experiments prove that the algorithm designed in this paper can find higher-quality samples.(3)The coupling matrix optimization problem in the computer-aided microwave filter tuning technology is a non-separable large-scale optimization problem.This paper designs a new cooperative coevolution strategy to decompose the problem,which is named as the multistage optimization strategy,and solve the sub-optimization problem in each stage withheuristic algorithm based on convergence speed controller.According to the test cases on the actual production line,the experiments prove that the algorithm designed in this paper can find the higher-accuracy coupling matrix and has universality.