Smooth norm limited non-negative matrix factorization clustering method based on graph regularization

Abstract

The invention provides a smooth norm limited non-negative matrix factorization clustering method based on graph regularization. The method comprises the steps of: S10, obtaining views to be clustered,and constructing a nearest neighbor graph; S20, constructing a target function of LP smooth norm limited non-negative matrix factorization based on graph regularization, wherein the target function comprises a smooth regularization item used for maintaining a geometric structure in a data space and improving smoothness and a graph regularization item used for marking sample category information;S30, by taking a Frobenius norm as a measurement standard, performing iterating for a preset number of times by using an iterative weighting method according to the target function to obtain the feature matrix of the to-be-clustered view; and S40, adopting a k-means clustering algorithm to analyze the feature matrix of each clustering view, so that view clustering can be realized. According to themethod, graph regular terms are added into NMF, and intrinsic geometrical information of a data set is respected while hidden semantics are found; the label information is taken as an additional hardconstraint to enable the marked samples in the high-dimensional space to have the same coordinates in the new low-dimensional space; and finally, an LP smoothing norm is added to improve the smoothness.

Description

technical field [0001] The invention relates to the technical field of data processing, in particular to a view clustering method. Background technique [0002] With the rapid development of communication technology and Internet technology, people's ability to acquire images is getting stronger and stronger, and the acquired images have shown the characteristics of high dimensionality, large amount of data, and complex scale. Although high-dimensional images will bring people a better visual experience, they also bring problems to the processing and storage of image data, so data dimensionality reduction has become a need. With the development of image processing technology to this day, researchers have proposed a variety of data dimensionality reduction methods according to different research methods. At present, the more commonly used data expression methods are principal component analysis (Principal Component Analysis, PCA), linear discriminant analysis (Linear Discrimin...

AI Technical Summary

Problems solved by technology

[0004] In view of the above problems, the present invention provides a clustering method based on graph regularization-based smooth norm restricted

Images