Parallel To Assist Double Population Differential Evolutionary Programming Based On Hadoop

Comparing with other algorithms, Evolutionary algorithms(Evolutionary Algorithms, EAs) has advantages in operation, versatility and efficiency.In recent years, more popular EAs branch with researchers is differential EAs, which is straight, easy-operation and easy-run and that has obviously advantages in solving optimal problems. However, the application of differential EAs is immature, which has many problems concerned researchers, for example, how to avoid it into a precocious convergence status in evolution, how to accelerate convergence speed. Now researchers optimize these problems with controlling parameter, selecting the appropriate differential strategy, integration of other optimization algorithms, double population evolution and populations evolution.This paper researches the improvement of differential EAs in double population evolution, which mainly concerns resolving complicated issues, and developing the efficiency in avoiding precocious convergence status in evolution, accelerate convergence speed. Details are as follows:1、 The question about differential EAs turning to premature convergence easily or convergence slowly is furthered which put forward a new evolutionary algorithm—DPADEP.The main idea of the DPADEP is : Double population evolution is divided into two stages: double population independent evolution phase and double population restructuring phase. In the double populations independent evolution stage,it introduced a assistance optimum mobility evolutionary model which configure the same assistance population for two equal divided subpopulation, while the subpopulation evolved parallelly with assistance population, and competed and replaced with each generation of assistance population. In double population restructuring phase, this paper improved optimum individual nonuniform evolutionary programming operator and proposed two new optimum individual evolutionary programming operator : optimum individual cauchy mutation operator evolutionary programming and optimum individual gaussian mutation operator evolutionary programming which were based on BGDEP, in order to replacing the worst individuals.2、 For differential evolutionary algorithm in solving high dimensional and complex issues of slow evolution of the parallel platform, this paper parallelled the assist differential evolutionary programming algorithm basing on the Hadoop. At the same time it contrasted the time of serial algorithm and parallel algorithm,which verified the effectiveness of parallel algorithm.