图卷积算法的研究进展  

A concise survey on graph convolutional networks

作　　者：郑睿刚 陈伟福[1,2] 冯国灿[1,2] ZHENG Ruigang;CHEN Weifu;FENG Guocan(School of Mathematics,Sun Yat-sen University,Guangzhou 510275,China;Guangdong Province key Laboratory of Computational Science,Sun Yat-sen University,Guangzhou 510275,China)

机构地区：[1]中山大学数学学院,广东广州510275 [2]中山大学广东省计算科学重点实验室,广东广州510275

出　　处：《中山大学学报（自然科学版）》2020年第2期1-14,共14页Acta Scientiarum Naturalium Universitatis Sunyatseni

基　　金：国家自然科学基金(61673018,61272338,61703443);广州市科技计划项目(201804010255)。

摘　　要：近年来,随着科学技术的发展,越来越多的数据以图的形式呈现和存储。图是不规则的数据,具有分散性和无序性,除了节点本身可赋予数据的特征外,边权信息更可以刻画节点间的相似性。虽然传统的深度卷积网络能有效处理图像、视频、语音等规则的数据,但直接用以处理图的数据效果并不理想。如何借鉴传统的卷积算法,提出适应图数据特点的学习算法,是当前深度学习研究的一个热点。文章拟对面向图数据的图卷积算法进行归纳总结,然而由于篇幅有限,无法对所有算法做到面面俱到的介绍,因此文章侧重于介绍模型背后的原理,分析并指出这些算法的优缺点,同时扼要介绍图卷积网络的主要应用。In recent years,many new technologies are constantly emerging in every aspect of our lives.More and more data have been generated and stored in graph format.Graphs are irregular data,which possess the characteristic of being distributive and disordered.Besides its capability that nodes can endow with data features,edge information can further depict the similarities among nodes.Despite the fact that classic convolutional neural networks are capable of handling regular format data such as images,videos and speech,directly applying these networks to graph data seems to be problematic.Recently,quite a few of researches were proposed to consider how to generalize classic convolutional neural networks for graph data and many high efficient learning algorithms were developed.This work aims to summarize and discuss the promising development of graph convolutional neural networks that were specifically designed for graph data.Nonetheless,due to the limited space,we cannot provide all the details of graph convolutional neural networks.Instead,we tend to introduce the motivations of those models,the analyses of the pros and cons of each model,and a brief summary of the major applications of graph convolutional neural networks.

关 键 词：图卷积神经网络 图的拉普拉斯矩阵 图的傅立叶变换 图的卷积变换 图的节点分类 图的分类 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]