Community classification method based on multi-dimensional graph convolutional neural network

Abstract

The invention discloses a community classification method based on a multi-dimensional graph convolutional neural network, and the method comprises the steps: carrying out the preprocessing of pre-extracted community graph data; constructing a multi-dimensional graph convolutional neural network model based on the K-dimensional relation matrix, the L-layer graph volume network and the full connection layer; and finally, calculating a cross entropy loss value according to an output result of the full connection layer and a standard classification result, feeding the cross entropy loss value back to the multi-dimensional graph convolutional neural network, and repeatedly training until the model is converged. According to the method, a new K-dimensional adjacency matrix is defined, and a new multi-dimensional graph convolutional network model is constructed, so that deep connection among community members can be found, and the training speed and prediction accuracy of the model are improved under the condition that the network depth and the data set scale are not increased.

Description

technical field [0001] The invention belongs to the field of deep learning, and in particular relates to a community classification method based on a multidimensional graph convolutional neural network. Background technique [0002] In recent years, convolutional neural networks have developed rapidly in many application directions, and have made breakthroughs in speech recognition, face recognition, general object recognition, motion analysis, natural language processing and even brain wave analysis. But for graph-structured data, many graph convolutional networks or recurrent neural networks cannot give a good solution. [0003] In real life, there are many, many irregular data structures, typically graph structures, or topological structures, such as social networks, chemical molecular structures, knowledge graphs, etc.; even languages ​​are actually complex tree structures inside The structure is also a graph structure; and like a picture, when doing target recognition,...

AI Technical Summary

Problems solved by technology

However, the existing algorithms are more or less slow in optimization, the classification accuracy drops seriously when the number of communities increases, and the deep relationship of the original data is not mined.

Images