Graph Neural Networks-based Multilabel Classification of Citation Network

Guillaume Lachaud, Patricia Conde-Cespedes

Abstract:

There is an increasing number of applications where data can
be represented as graphs. Besides, it is well-known that artificial intelli-
gence approaches have become a very active and promising research field,
mostly due to deep learning technologies. However popular deep learning
architectures were designed to treat mostly image and text data. Graph
Neural Network is the branch of machine learning which builds neural
networks for graph data. In this context, many authors have recently
proposed to adapt existing approaches to graphs and networks. In this
paper we train three models of Graph Neural Networks on an academic
citation network of Computer Science papers, and we explore the advan-
tages of turning the problem into a multilabel classification problem.