Particle Swarm Optimization Algorithm Based On Extended P Systems And Its Application In Data Clustering

PSO algorithm is a kind of bionic intelligence optimization algorithm,and the computational model is abstracted from the foraging behavior of the birds in nature.In PSO,each particle adjusts the movements according to individual and social experience,so that the complex behavior of the population can be finished by simple behavior of particles.However,because of the uncertainty and randomness of the swarm intelligence algorithm,PSO is easily trapped into local optima and appeared premature phenomenon at the end of the computation process.Especially for solving large dimensional optimization problems,the convergence precision and speed of PSO need to be further improved and optimized.Membrane computing,also known as P systems,is a new computational model which simulates the biochemical reactions of cells,tissues and organs.It is an important branch of natural computing,the computing models mainly consist of membrane structure,objects and rules.Due to the independency of cells,tissues and organs,P systems run in the distributed parallel model.Many researches show that some models of P systems present the same computing power as Turing machines and is more efficient to some extent.Data clustering is a hot issue in the data mining area,especially in the explosive growth of mobile data.Traditional PSO clustering algorithm has some limitations in solving this problem.Therefore,it is very necessary to introduce the new computing methods or calculation models in order to improve the clustering performance of PSO.Image segmentation has always been an important application of clustering method.How to improve the quality of segmentation results and reduce the computational complexity of PSO algorithm are hot topics in the studies of image segmentation.The main research contents of this thesis are as follows:(1)The design of three computing modes of extended P systemsThree kinds of extended membrane computing model are proposed which based on the traditional P system,including Evolution communication cell-like P system with channels states and membrane division and dissolution(ECP),Evolution communication tissue-like P system with promoter and inhibitor(ETP)and Chained P system(CP).These computing models are constructed by the structures,objects and rules.At last,the calculation ability of the three extended models is introduced to verify the effectiveness.(2)A multiple particle swarm coevolution model based on extended P system is proposed(PSO-P)A multiple particle swarm coevolution model based on extended P system is proposed,and named PSO-P.The distributed computing framework and evolutioncommunication rules of P systems are introduced to this coevolution model.The particle population is divided into many sub-populations,and the co-evolution process is completed by the evolution-communication mechanism in the sub-populations.(3)A dynamic PSO inspired by starling flock and ECP system is proposed(DSPSO-ECP)DSPSO-ECP algorithm is proposed,which combined with improved PSO inspired by starling flock and ECP system.In DSPSO-ECP,the Fitness-Euclidean ratio method is used to select the neighbors of particle,and the position of particle is determined by the individual and global information.The rules of membrane division and dissolutions in the ECP system is to introduced to generate and dissolve the sub-population to escape local optima.(4)A cooperative evolutionary QPSO based on ETP system is proposed(CQPSOETP)CQPSO-ETP algorithm is proposed,which combined with the improved QPSO algorithm based on the cooperative learning and self-adaptive adjustment mechanism and ETP system.In CQPSO-ETP,the population diversity function and cooperative learning are introduced,and the evolution rules with promoters and inhibitors are defined respectively.The information among the sub-population is used to dynamically adjust the evolution/computation process of particles.(5)A PSO with environmental factor inspired k-means and complex CP system is proposed(KEPSO-CCP)In the traditional PSO clustering algorithm,the clustering information is introduced as the environmental factor,and the information of individual,population and environment are determined the flying direction of each particle in the next iteration.At the same time,the crossover and mutation rules in the complex chained P system based on differential evolution algorithm is used to guide the local search for the global optima from the sub-population.(6)Three kinds of segmentation methods based on PSO and extended P systems are proposedThree kinds of image segmentation algorithms are proposed,including DSPSOECP,KEPSO-CCP and CQPSO-ETP.The first hybrid PSO algorithm combined to FCM clustering method to solve the gray image segmentation problems.The second PSO clustering algorithm uses the maximum inter-class variance as the objective function to solve the multi-level threshold gray image segmentation problems.The third QPSO segmentation consists of two clustering algorithm,CQPSO-ETP and SLIC,which is used for solving the color image segmentation problems.In summary,this paper mainly proposes three computational models of extended P system,and builds a multi-population co-evolution model based on particle swarm optimization and extended P system.Meanwhile,the PSO,QPSO and PSO clustering algorithms are introduced into the computational framework of these three extended P systems,which combined with the basic concepts of membrane division/membrane dissolution,promoter/inhibitor,chained structure in P system.Then three PSO algorithms based on the extended P system are proposed,which combined with optimized and clustering model to the image segmentation problem in order to improve the final segmentation effect of the PSO segmentation algorithms.