A Graphical Model Based on Multi-view Dictionary Learning

Abstract

The invention discloses a graph model based on multi-view dictionary learning, uses principal component analysis and linear discriminant analysis to perform data dimensionality reduction and preprocessing on original graph data, removes redundant features in the data while retaining high discrimination of the data Then use the multi-view dictionary learning method to learn the essential features contained in the data, and train a comprehensive dictionary, an analysis dictionary, a sparse code corresponding to the sample, and a SVM linear classifier; then input the sparse code of the sample SVM classifier, after the classifier is processed, a plurality of prediction labels under different perspectives are generated according to the multi-view principle; the predicted multi-view labels are synthesized by a voting mechanism, and the final sample labels are generated and used for the calculation of model accuracy; the present invention has It has the advantages of high information utilization rate, more efficient decision-making, and strong specificity.

Description

A Graph Model Based on Multi-View Dictionary Learning technical field The present invention relates to graph mining and dictionary learning technical field, be specifically related to a kind of graph based on multi-view dictionary learning Model. Background technique In recent years, the development of deep learning has made major breakthroughs, and is widely used in language recognition, target detection, machine learning machine translation and other fields, and demonstrated its powerful feature extraction capabilities. Deep learning is already in Euclidean data, as shown in the figure film, video, and voice, and have achieved great success because such data have some good properties, such as translation invariance, Partial connectivity and semantic compositionality of image data, but data generated from non-Euclidean domains, such as Graph data and Manifold data, but does not have the above-mentioned good properties, and often has a very complex structure, such...

AI Technical Summary

Problems solved by technology

Dictionary learning is widely used in image denoising, super-resolution, compressed sensing and other fields, but so far, there are few or few studies on applying multi-view technology to dictionary learning, and applying dictionary learning to graph data

For the algorithm, this is a vacancy in application and a lack of function; at the same time, the existing technology has shortcomings and deficiencies such as low information utilization rate, lack of function, high consumption of computing resources, and long training period.

Images