A Multi-view Data Clustering Method Based on Interregularization Constrained Subspace Expression

Abstract

The invention discloses a multi-view data clustering method based on mutual regularization constrained subspace expression. The method comprises the following steps: forming a multi-view database, extracting view characteristics of different view data; selecting a mutual regularization method, and determining the mutual regularization optimization constraints; based on an optimization objective function and the perspective characteristics of different perspective data, the collaborative representation vectors of all samples are obtained; the cooperative representation vectors are sorted according to the order of their corresponding samples to obtain the subspace expression matrix; the subspace expression matrix is ​​processed , to obtain the affinity matrix; divide the affinity matrix according to the number of clusters required to obtain the sample clustering results of the multi-view database. The present invention uses subspace expression to mine the implicit structural information between samples, and uses two mutual regularization methods to constrain the subspace expression of different perspectives, so as to utilize the complementary information of multiple perspectives to further strengthen the accuracy of the implicit structural information of the sample set. Representation, which can be widely used in multi-view data clustering.

Description

technical field [0001] The invention relates to the technical field of pattern recognition, in particular to a multi-view data clustering method based on mutual regularization constraint subspace expression. Background technique [0002] Data in the real world often have multiple perspectives. The perspectives here can be different feature expressions of the same image, or data of different modalities. For example, web page information includes not only image information, but also text information and hyperlink information. Multi-view clustering, as a basic task of pattern recognition, aims to use complementary information from different perspectives to improve clustering performance. (2) Using complementary information provided by multi-view data. When traditional multi-view clustering methods solve these two challenges, they generally find a unified and discriminative low-dimensional representation of multi-view data, but they cannot explicitly utilize the prior informati...

AI Technical Summary

Problems solved by technology

When traditional multi-view clustering methods solve these two challenges, they generally find a unified and discriminative low-dimensional representation of multi-view data, but they cannot explicitly use the prior information of the data itself, such as sparsity, Collaborative Representation Features

Therefore, traditional multi-view clustering methods cannot effectively cluster multi-view data.

Images