Graph Deep Learning: State of the Art and Challenges

The last half-decade has seen a surge in deep learning research on irregular domains and efforts to extend convolutional neural networks (CNNs) work irregularly structured data. graph emerged as particularly useful geometrical object learning, able represent variety of well. Graphs can various complex systems, from molecular structure, computer social traffic networks. Consequent the extension CNNs graphs, great amount been published that improves inferential power computational efficiency graph-based (GCNNs). is incipient, however, our understanding relatively rudimentary. majority GCNNs are designed operate with certain properties. In this survey we review state representation perspective learning. We consider challenges have neglected work, largely because numerous theoretical difficulties they present. identify four major learning: dynamic evolving edge signals information, estimation, generalization models. For each problem discuss practical issues, relevant research, while highlighting limitations art. Advances these would permit be applied wider range domains, situations where models previously limited owing obstructions applying model domains' natures.