Research Of Quantum Affinity Propagation Clustering Algorithm

Affinity propagation clustering algorithm is an unsupervised clustering algorithm in recent years,but there are two shortcomings in practical application:(1)The AP algorithm considers all data sample points as candidate cluster centers,and the same values are used to describe the probability of each data sample as clustering center by the preference,because it can not be distinguished.The data samples become the difference of clustering centers,so the algorithm iterates too many times and the computation efficiency of clustering is low.(2)The AP algorithm has a better clustering effect when facing the cluster of data sets,but the data sample differentiation ability is weak and the clustering results are poor.In view of the above two shortcomings of AP algorithm,this paper proposes a quantum AP clustering algorithm with local density optimization.The main research work is as follows.1.In view of the above problem(1),we use quantum evolutionary algorithm to optimize the preference of AP algorithm,and find out the probability of each data sample as clustering center through quantum search,so as to realize the heuristic prior knowledge of clustering center in the preference.First,the preference are encoded by the quantum superposition state,and the quantum superposition state is searched by the quantum rotation gate.The preference of the near worry is found out for AP algorithm clustering,and the adaptive optimization of AP preference is achieved.In order to solve the problem of poor identification of structured data by AP algorithm,the Gauss kernel function is used instead of Euclidean distance as a similarity measure to map data from low dimensional space to high dimensional space to improve the recognition ability of the algorithm for complex data.2.Aiming at the above problem(2),the local density optimization strategy is introduced to mark the data points with large sparsity,and the density relationship is redivided after the end of clustering to distinguish the distribution density of data samples.Finally,the proposed algorithm is verified on the UCI data set.The experimental results show that the clustering accuracy and clustering efficiency of the proposed algorithm are superior to those of similar algorithms,and the idea of improving the AP algorithm is feasible.