Motif-based Graph Self-Supervised Learning for Molecular Property Prediction

To bridge this gap, we propose Motif-based Graph Self-supervised Learning (MGSSL) by introducing a novel self-supervised motif generation framework for GNNs.