Abstract

AbstractWith the ability of representing structures and complex relationships between data, graph learning is widely applied in many fields. The problem of graph classification is important in graph analysis and learning. There are many popular graph classification methods based on substructures such as graph kernels or ones based on frequent subgraph mining. Graph kernels use handcraft features, hence it is so poor generalization. The process of frequent subgraph mining is NP-complete because we need to test isomorphism subgraph, so methods based on frequent subgraph mining are ineffective. To address this limitation, in this work, we proposed novel graph classification via graph structure learning, which automatically learns hidden representations of substructures. Inspired by doc2vec, a successful and efficient model in Natural Language Processing, graph embedding uses rooted subgraph and topological features to learn representations of graphs. Then, we can easily build a Machine Learning model to classify them. We demonstrate our method on several benchmark datasets in comparison with state-of-the-art baselines and show its advantages for classification tasks.KeywordsGraph classificationGraph miningGraph embedding

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call