IF-001
Dual Graph Convolutional Networks for Graph-Based Semi-Supervised Classification

◎Chenyi Zhuang（AIST）・Qiang Ma（Kyoto Univ.）

How to make computers sufficiently understand a complex graph is an important task in a range of different fields. For instances, in the fields of the Internet, social networks, biological networks, and many others, more and more structured data is becoming available. As a result, it is interesting and necessary to devise advanced methodologies to extract meaningful data from these various graphs. In this paper, we present a scalable graph convolutional networks method for graph-structured data analysis, and then apply it to solve the graph-based semi-supervised classification problem. To make computers sufficiently understand a graph, we proposed a dual graph convolutional networks method that performs graph convolution from two different views of the raw graph: (1) local-consistency-based view and (2) global-consistency-based view.