Open AccessRG-GCN: Improved Graph Convolution Neural Network Algorithm Based on Rough Graph
Author(s) -
Weiping Ding,
Bairu Pan,
Hengrong Ju,
Jiashuang Huang,
Chun Cheng,
Xinjie Shen,
Yu Geng,
Tao Hou
Publication year - 2022
Publication title -
ieee access
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.587
H-Index - 127
ISSN - 2169-3536
DOI - 10.1109/access.2022.3198730
Subject(s) - aerospace , bioengineering , communication, networking and broadcast technologies , components, circuits, devices and systems , computing and processing , engineered materials, dielectrics and plasmas , engineering profession , fields, waves and electromagnetics , general topics for engineers , geoscience , nuclear engineering , photonics and electrooptics , power, energy and industry applications , robotics and control systems , signal processing and analysis , transportation
The graph convolution neural network uses topological graph to portray inter-node relationships and update node features. However, the traditional topological graph can only describe the certain relationship between nodes (that is, the weight of the connecting edge is a fixed value), while ignoring the uncertainty widely existing in the real world. These uncertainties not only affect the relationship between nodes, but also affect the final classification performance of the model. In order to overcome this defect, a graph convolution neural network algorithm based on rough graph is proposed in this paper. Specifically, the algorithm first constructs a rough graph using a combination of the upper and lower approximation theory of the rough set and the edge theory of the topological graph, the paired maximum-minimum relationship values are used to characterize the uncertain relationship between nodes. Then, this paper designs an end-to-end training neural network architecture based on rough graph, the trained rough graph is fed to this neural network to update node features with these uncertain relationship. Finally, nodes are classified according to these learned node features. The experimental results on real data show that the proposed algorithm can significantly improve the accuracy of node classification compared with the traditional graph convolution neural network.