Self-supervised learning is currently gaining a lot of attention, as it\nallows neural networks to learn robust representations from large quantities of\nunlabeled data. Additionally, multi-task learning can further improve\nrepresentation learning by training networks simultaneously on related tasks,\nleading to significant performance improvements. In this paper, we propose\nthree novel self-supervised auxiliary tasks to train graph-based neural network\nmodels in a multi-task fashion. Since Graph Convolutional Networks are among\nthe most promising approaches for capturing relationships among structured data\npoints, we use them as a building block to achieve competitive results on\nstandard semi-supervised graph classification tasks.\n