Image clustering method, system, device and medium based on discrete multi-view clustering

Abstract

The disclosure discloses an image clustering method, system, device and medium based on discrete multi-view clustering, including: learning an ideal structured graph for multi-view data; The embedding of the original data retains the flow structure of the original data; at the same time, a mapping matrix is ​​learned to solve the clustering problem of out-of-sample data; the objective function is solved, the continuous labels are discretized, and the final discrete cluster labels are obtained; in the whole learning process In both, considering the complementarity of multi-view data and the corresponding importance of different view features to learning tasks, this method automatically assigns appropriate weights to each view feature. The invention improves the accuracy of multi-view clustering.

Description

technical field [0001] The present disclosure relates to the technical field of image clustering, in particular to an image clustering method, system, device and medium based on discrete multi-view clustering. Background technique [0002] The statements in this section merely mention background art related to the present disclosure and do not necessarily constitute prior art. [0003] In the process of realizing the present disclosure, the inventors found that the following technical problems existed in the prior art: [0004] With the in-depth development and application of information technology, the era of big data has arrived. In order to describe data information more accurately and completely, multi-view data appears and is widely used in different research fields, such as data mining, machine learning, and computer vision. The same thing can be described in many different ways or from different angles, and these multiple descriptions constitute multiple views of th...

AI Technical Summary

Problems solved by technology

Since the K-means algorithm needs to initialize the centroid randomly, the clustering result is always unstable

In addition, clustering methods all face a practical problem, they cannot solve the problem of clustering samples outside the training set

[0006] Although many graph-based multi-view clustering methods have been proposed, the existing methods still have several shortcomings mentioned above, and the methods need to be improved.

Images