awesome-unsupervised-gnns
List of unsupervised (self-supervised) graph neural network (GNN) methods. The purpose of this repository is to provide a literature survey of self-supervision signals (loss functions) that can be used to train GNNs in an unsupervised manner.
Node related tasks
1. Constrastive based self-supervision
- [ARXIV 2020] Sub-graph Contrast for Scalable Self-Supervised Graph Representation Learning [paper]
- [ARXIV 2020] Graph InfoClust: Leveraging cluster-level node information for unsupervised graph representation learning [paper]
- [ICML 2020] Contrastive Multi-View Representation Learning on Graphs [paper] [code]
- [WWW 2020] Graph Representation Learning via Graphical Mutual Information Maximization [paper] [code]
- [ICLR 2019] Deep Graph Infomax [paper] [code]
- [NIPS 2017] Inductive Representation Learning on Large Graphs [paper] [code]
2. Reconstrunction based self-supervision
- [WWW 2020] Graph Representation Learning via Graphical Mutual Information Maximization [paper] [code]
- [ARXIV 2019] Graph Attention Auto-Encoders [paper] [code]
- [ICCV 2019] Symmetric Graph Convolutional Autoencoder for Unsupervised Graph Representation Learning [paper] [code]
- [IEEE on Cybernetics 2019] Learning Graph Embedding with Adversarial Training Methods [paper]
- [IJCAI 2018] Adversarially Regularized Graph Autoencoder for Graph Embedding [paper] [code]
- [NIPS Workshop 2016] Variational Graph Auto-Encoders [paper] [code]
3. Adversarial self-supervision
- [IJCAI 2020] Adversarial Mutual Information Learning for Network Embedding [paper]
- [IEEE on Cybernetics 2019] Learning Graph Embedding with Adversarial Training Methods [paper]
- [IJCAI 2018] Adversarially Regularized Graph Autoencoder for Graph Embedding [paper] [code]
4. Other
- Interactive Clustering
- Hop Count Prediction
- [ARXIV 2020] Self-Supervised Graph Representation Learning via Global Context Prediction [paper]