Reliable multi-view learning method and device based on data reconstruction

Abstract

The embodiment of the invention provides a reliable multi-view learning method and device based on data reconstruction, and belongs to the technical field of image clustering. Low-rank self-representation is introduced to construct a data reconstruction model so as to make up for the damage of noise to the real structure of data; in order to better explore the two-dimensional similarity between different views, a low-rank tensor is adopted to capture the high-order correlation between similarity images; a combined embedded framework is developed, a data reconstruction model, similarity graph learning and a low-rank tensor model are put into the framework, and the adaptability and robustness of the method are promoted; in addition, an effective numerical function solving method is designed to obtain an optimal value of a variable coupled in a target function; compared with an existing excellent method, the invention is higher in accuracy and more stable in performance.

Description

technical field [0001] Embodiments of the present invention relate to the field of image clustering, and more specifically, embodiments of the present invention relate to a reliable multi-view learning method and device based on data reconstruction. Background technique [0002] With the rapid development of science and technology, the information age has come quietly. The data that can be obtained every day shows an exponential growth, which makes most of the obtained data unlabeled. Therefore, clustering technology highlights its unique value. Among many applications of clustering technology, since most of the data obtained in the real world are presented in the form of multi-view, multi-view clustering method has become a hot spot nowadays. However, most multi-view clustering methods focus on the construction strategy or mechanism, thus ignoring the influence of noise on the real distribution structure of the data, so that the obtained fusion map cannot truly reflect th...

AI Technical Summary

Problems solved by technology

However, most of the multi-view clustering methods focus on the construction strategy or mechanism, thus ignoring the influence of noise on the real distribution structure of the data, so that the obtained fusion map cannot truly reflect the distribution of the data.

Images