Study On Natural Computation Method Based On Dimension Attribute Strategy

With the rapid development of information age,natural computing method has become a crucial part of the process of social intelligence.When many complex things and phenomena appear,the multivariable data,called high-dimensional data,will be derived.In the field of natural computing,with the continuous improvement of dimension,the information provided by data will be more abundant and comprehensive.The algorithm optimization ability is improved,and it also faces the problems of dimension disaster,large computation amount and poor time performance.In view of these problems,this paper proposes a method based on dimension attribute strategy to solve the above problems.The research results of this paper are as follows:A natural computing strategy based on dimension space segmentation search is proposed.The strategy mainly uses the grouping method of three-dimensional space as a group,which makes the high-dimensional space directly mapped to the visual 3D space rectangular coordinate system,which makes the complex and unobservable problem become observable problem.At the same time,the number of the individual after the space segmentation is divided into individual,which is based on reducing the dimension Indirectly,it increases the size of individuals,makes individuals distribute in a wider search space,and effectively increases population diversity.The algorithm iterates to a certain extent,and can synthesize the individual into the original individual by number index.By calculating the fitness value,some of the poor individuals can be deleted,and then the time performance can be balanced.Finally,the optimal individual output fitness value can be synthesized by searching the global optimal position of individual in the group by number index,which makes the algorithm have better optimization ability.At the same time,the convergence of the strategy is analyzed by using Markov chain.This paper proposes a natural computing strategy based on the idea of local linear embedding dimension reduction.By calculating the local reconstruction weight matrix,the reconstruction error is minimized,and the weight matrix is used for low dimensional embedding,and the computing eigenvector is used to extract the minimum eigenvector to achieve the dimension reduction.At the same time,the selection of neighbor particles in the algorithm is adopted to minimize the reconstruction error It can speed up the running time,enhance the global search ability and ensure the performance of the algorithm.The above two dimensional attribute strategies not only balance the effectiveness of the algorithm,but also realize that the algorithm process is not dependent on the specific steps of the algorithm,which can be applied to multi-population intelligent algorithms in the natural field.The proposed two strategies are applied to mainstream intelligent algorithms,such as particle swarm optimization(PSO),genetic algorithm(GA),cuckoo algorithm(CUCO)and differential evolution algorithm(DLE),and their performance is verified in standard test functions.At the same time,they are compared with several mainstream algorithms at present.Experiments show that the proposed strategy can improve the convergence speed and optimization ability obviously.