Graph-based neural network models with multiple self-supervised auxiliary tasks

• Graph-based neural network models exploiting multiple self-supervised auxiliary tasks. We propose three new tasks for graph-based networks. Vertex features autoencoding. Corrupted vertex reconstruction. embeddings Self-supervised learning is currently gaining a lot of attention, as it allows networks to learn robust representations from large quantities unlabeled data. Additionally, multi-task can further improve representation by training simultaneously on related tasks, leading significant performance improvements. In this paper, we novel train in fashion. Since Graph Convolutional Networks are among the most promising approaches capturing relationships structured data points, use them building block achieve competitive results standard semi-supervised graph classification