1

We present Topological Augmentation (TopoAug), a novel graph augmentation method incorporating higher-order node relations for real-world complex networks, and provide 23 novel real-world graph datasets across various domains.

We propose an approach to improve GNN-based graph generative models including GCPN, GraphAF and GraphEBM, and investigate the correlation between GNN expressiveness in the graph prediction and graph generation contexts.

We propose GraphAC (Graph Adversarial Collaboration), a conceptually novel, principled, task-agnostic, and stable framework for evaluating GNNs through contrastive self-supervision, without the need of handcrafted augmentations.

We present Meowjong, an AI for 3-player Mahjong (Sanma) using deep reinforcement learning. We define an informative and compact 2-dimensional data structure for encoding the observable information in a Sanma game. We pre-train 5 CNNs for Sanma’s 5 actions, and enhance the major action’s model via self-play RL using the Monte Carlo policy gradient method.