| With the rapid development of Computer Application Technology,efficient optimization techniques and intelligent computing technology has put forward higher requirements to update and has been used in a variety of engineering problems.In view of practical engineering difficulties,such as complexity,constrain, non-linear,many local minima and difficulties in modeling features,a new type of intelligent optimization methods which meet the practical engineering need has become an important research direction.Particle Swarm Optimization as an emerging smart computing technology has brought more and more researchers' focus.It has strong relationship with artificial life,especially with the evolution strategy and genetic algorithm.In the absence of centralized control and need not to provide the overall information,Swarm Interlligence has provided the foundation to look for complex distributed solutions.This paper described and analyzed the particle swarm algorithm in cluster analysis and multi-constraint multicast routing problem application.In cluster analysis,the paper proposes PSO clustering algorithm with space theory and chaotic characteristics,and base on its,the paper also extends the algorithm to give a general algorithm to solve clustering with no specific cluster numbers;in the multi-constrained multicast routing problem,unlike traditional methods,the algorithm use tree transformation method,relying on the particle swarm theory to solving the multicast routing problem(1) At present,the clustering algorithm can be divided into the following categories:patition methods,hierarchical methods,density-based methods and grid-based methods.Many ways need people to provide some constrains,such as configure the number of cluster,and cluster results are offen very sensitive with the initial state and algorithm parameters.K-Means clustering algorithm is the representation of such a clustering algorithm,the final clustering effect by the initial cluster center of great influence,the result is often the local optimal solution,in order to obtain better clustering results,computational complexity increases rapidly. And a stable and can converge to global optimal algorithm is the actual industry needs.This paper based on PSO algorithm proposes the SPA(Spatial PSO Algorithm) algorithm,in order to ensure that the data in the same cluster are similar and the data in the different clusters are unsimilar,the algorithm introduces spatial traits in the particle.At the same time,in order to solve the PSO algorithm precocious questions, it involves the chaotic idea.Simulation results show that the algorithm does well in digital properties cluster problem through the introduction of interior space features and chaos of thought,the average performance has also been significantly improved.(2)As for K-Means clustering algorithm,the algorithm need to specify the number of clustering problem in the begining,the paper involves the Bayesian information decision idea in the algorithm,and put forward the X-SPA(X-Spatial PSO Algorithm) clustering algorithm The X-SPA algorithm can adjust and calculate the real cluster number in the process of running algorithm The experimental results also show that the algorithm not only give a more accurate number of cluster categories and also a high accuracy of clustering.(3)In this paper,we propose a tree-based Particle Swarm Optimization PSOTREE algorithm to solve multicast routing problem.Unliking traditional algorithms which usually find every road from the source node to every destination node and then convert them to a multicast tree,PSOTREE algorithm use a pattern which directly calculate multicast tree.PSO-based tree algorithm has simplified the search tree mechanism,and the algorithm efficiency also has been improved. Simulation results show that compared with traditional algorithms,PSOTREE algorithm can faster converge to the approximate optimal solution,enhance the efficiency,the more the number of group members,the more obvious advantages. |
PDF Full Text Request