The Immune Genetic Algorithm And The Research Of Its Application Based On Parallel

Along with the in-depth application of genetic algorithms, and as any of the genetic algorithm optimization algorithm has other advantages, therefore, Genetic Algorithm for the Optimization Calculation of a wide range of applications.but people have found that for a variety of problems the genetic algorithm will create "premature convergence" or weak ability of local search,which affects the algorithm to overall search for the optimal solution. With the continuous development of science and technology, the continued expansion of the scale of the problem, and facing increasing complexity of the search space, genetic algorithms for optimizing the efficiency and quality have become "too pale"! To speed up the timeliness and accuracy of decision-making,we can research and test GAs as the example as the the distribution location of OBD and OUN of Passive Optical Network on Cluster Of Workstations(COW) in this paper.Firstly,based on the mechanism of such features as antigen recognition,variability of antibody and immune memory in immune systems,a new improved GA,namely,Immune Genetic Algorithm is presented in this thesis.In order to overcome premature convergence and find out optimal solution,the immune mechanism of creature is used in IGA and antibodies will be promoted or restrained according to the computation result of affinity between antibodies,which reserves the excellent antibodies as well as guarantees the variability of antibody.In addition,the IGA'slocal searching ability is improved by combining it with gradient method.The computation results show that both global and local searching abilities of the IGA are improved and premature convergence is overcome availably.The effectiveness and the superiority of IGA is proved by optimization experiments using another optimization algorithm comparisons to.Secondly,this paper summarizes the development and feature of parallel GA.We introduce hardware system and of parallel process and software in the parallel environment—MPI on COW. Thirdly,the point is this:we analyze the inherent potential parallel in Parallel Genetic Algorithm.In terms of master-slave parallel programming design,we put forward a Parallel immune Genetic Algorithm of shortest path based on MPI in COW.So this can avoid the premature convergence problem and enhance the global convergence.In the process of parallel algorithm design including partition,communication,combination and mapping,put forward the partition principle of genetic algorithm used to originate the groups.Finally in the experiment part by MPI, we configure COW parallel environment and in the platform of Windows and MPI,we program the algorithm with Visual C++6.0.The experiment platform and Parallel Genetic Algorithm,and establishes the foundation for Emergency Decision System.By analyzing and comparing to experiment data,we compute the accelerated performance.Results show The algorithm has high fitness and optimized speed.But the size of this parallel algorithm is not enough.Genetic parameter is not allowed to modify,so is the content and time of message passing.What we can do is making it more perfect.