Method for realizing and classifying bilinear graph neural network model for modeling neighbor interaction

Abstract

The invention discloses a method for realizing and classifying a bilinear graph neural network model for modeling neighbor interaction. The method comprises the following steps: performing weighted average on neighbor node feature information by using a linear aggregator; modeling interaction between neighbor nodes, or between a target node and each neighbor node, or between all nodes by using a bilinear aggregator, and averaging the interacted information; combining the results obtained in the first two steps by adopting a linear combination mode so as to construct a bilinear graph neural network model and obtain a feature representation vector of the target node. According to the method, more effective node representation can be obtained, so that the graph node classification accuracy isimproved.

Description

technical field [0001] The invention relates to the technical field of machine learning and graph data mining, and in particular to an implementation method and classification method of a bilinear graph neural network model for modeling neighbor interaction. Background technique [0002] The graph neural network learns the representation of the nodes in the graph by performing convolution operations on the data based on the graph structure, and then applies it to many fields such as social science, natural language processing, computer vision, and recommendation systems. [0003] The frequency domain (Spectral) graph neural network performs convolution operations on the representation of nodes in the Fourier domain, and usually needs to define the eigenvectors of the graph Laplacian matrix as a Fourier basis. This process requires eigendecomposition of the matrix, and the computational complexity is very high. In order to improve efficiency, the K-degree Chebyshev polynomia...

AI Technical Summary

Problems solved by technology

Therefore, the traditional graph neural network will cause insufficient aggregated node information due to these two reasons, so that the node representation cannot be effectively learned, and the classification accuracy will be reduced.

Images