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.