Tranform-based quotation network classification model establishment and classification of graph convolution network

Abstract

The invention discloses a Transform-based graph convolutional network citation network classification model establishment method, which comprises the following steps of: firstly, acquiring citation network data which comprises the steps of determining subject identities (papers and authors) of nodes, collecting corpus features of the nodes, determining labels of the nodes and determining relationships among the nodes, and then establishing a Transform-based graph convolutional network model, comprising a K-layer simplified graph convolutional network module and a transformed Transform encoder, the method comprises the steps that firstly, a simplified graph convolution network is used for conducting feature convolution propagation on all nodes, a Transform encoder is used for learning a global feature for each layer of features of all the nodes of a training set to be used for classification, and finally the trained Transform encoder is used for classifying test nodes.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence, and in particular relates to a method for establishing and classifying a citation network classification model based on a Transformer graph convolution network. Background technique [0002] Convolutional Neural Networks (CNN) have been widely used in computer vision and have achieved excellent performance, especially for data with Euclidean features such as images. The convolutional layer in CNN learns a variety of different local filters. , and extract high-level features from the image by filtering. Then it is very important to effectively extract features on the data of the relational structure of graph. Analogy to the convolution operation on the image and the graph signal processing, there are two definitions of graph convolution. One is defined in the spectral domain, such as ChebNet, GCN, SGC. The other is defined in the spatial domain, such as GarphSage, GAT. [0003] ...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the technical problem that the deep learning of the convolutional network cannot be performed in the prior art, the purpose of the present invention is to provide a citation network classification model establishment and classification method based on the Transformer-based graph convolutional network

Images