Awesome-Graph-Contrastive-Learning

Collection of resources related with Graph Contrastive Learning.

Contents

• Papers
  • 2019
  • 2020
  • 2021
• Library

Papers

2019

1. Deep Graph Infomax Petar Veličković, William Fedus, William L. Hamilton, Pietro Liò, Yoshua Bengio, & R Devon Hjelm. ICLR 2019.
   • Paper: OpenReview | Code: PyTorch, tf_geometric
   • Method(DGI): Local-Global Mutual Information Maximization
   • Experiment:
     • Task: Transductive Node Classification | Datasets: Cora, Citeseer, Pubmed. | Baselines: Raw features, Label Propagation, DeepWalk, DeepWalk + features, GCN, Planetoid.
     • Task: Inductive Node Classification | Datasets: Reddit, PPI. | Baselines: Raw features; DeepWalk; DeepWalk + features; GraphSAGE-GCN; GraphSAGE-mean; GraphSAGE-LSTM; GraphSAGE-pool; FastGCN; Avg. pooling.

2020

1. InfoGraph: Unsupervised and Semi-supervised Graph-Level Representation Learning via Mutual Information Maximization Fan-Yun Sun, Jordan Hoffmann, Vikas Verma, Jian Tang. ICLR 2020.

   • Paper: OpenReview | Code: PyTorch
   • Method: Local-Global Mutual Information Maximization (Batch-wise Negative Sampling, Multi-scale), GIN
   • Experiment:
     • Task: Graph Classification | Datasets: MUTAG, PTC-MR, RDT-B, RDT-M5K, IMDB-B, IMDB-M. | Baselines: RandomWalk, Shortest Path Kernel, Graphlet Kernel, Weisfeiler-Lehman Sub-tree Kernel, Deep Graph Kernels, Multi-Scale Laplacian Kernel, node2vec, sub2vec, graph2vec.
     • Task: Semi-supervised Molecular Property Prediction | Datasets: QM9. | Baselines: Mean-Teachers.

2. Contrastive Multi-View Representation Learning on Graphs Kaveh Hassani, Amir Hosein Khas Ahmadi. ICML 2020.

   • Paper: MLR | Code: PyTorch
   • Method(MVGRL): Multi-View Local-Global Mutual Information Maximization
   • Experiment:
     • Task: Node Classification | Datasets: Cora, Citeseer, Pubmed. | Baselines: MLP, Iterative Classification Algorithm, Label Propagation, ManiReg, SemiEmb, Planetoid, Chebyshev, GCN, MoNet, JKNet, GAT, Linear, DeepWalk, GAE, VERSE, DGI.
     • Task: Node Clustering | Datasets: Cora, Citeseer, Pubmed. | Baselines: VGAE, MGAE, ARGA, ARVGA, GALA.
     • Task: Graph Classification | Datasets: MUTAG, PTC-MR, IMDB-BIN, IMDB-MULTI, REDDIT-BIN. | Baselines: Shortest Path Kernel, Graphlet Kernel, Weisfeiler-Lehman Sub-tree Kernel, Deep Graph Kernels, Multi-scale Laplacian Kernel, GraphSAGE, GIN-0, GIN-ε, GAT, RandomWalk, node2vec, sub2vec, graph2vec, InfoGraph.

3. GCC: Graph Contrastive Coding for Graph Neural Network Pre-Training Jiezhong Qiu, Qibin Chen, Yuxiao Dong, Jing Zhang, Hongxia Yang, M. Ding, Kuansan Wang, Jie Tang. KDD 2020.

   • Paper: ACM, Arxiv | Code: DGL

4. Graph Contrastive Learning with Augmentations Yuning You, Tianlong Chen, Yongduo Sui, Ting Chen, Zhangyang Wang, Yang Shen. NeurIPS 2020.

   • Paper: NeurIPS, Arxiv | Code: torch_geometric

5. Unsupervised Graph Representation by Periphery and Hierarchical Information Maximization Sambaran Bandyopadhyay, Manasvi Aggarwal, M. Murty. Arxiv 2020.

   • Paper: Arxiv
   • Method(GraPHmax): Periphery Information Maximization, Hierarchical Information Maximization
   • Experiment:
     • Task: Graph Classification | Datasets: MUTAG, PTC, PROTEINS, NCI1 and NCI09, IMDB-BINARY, IMDB-MULTI. | Baselines: Graphlet Kernel, RandomWalk Graph Kernel, Propagation Kernels, Weisfeiler-lehman Graph Kernels, AWE-DD, AWE-FB, DGCNN, PSCN, DCNN, ECC, DGK, DiffPool, IGN, GIN, 1-2-3GNN, 3WL-GNN, node2vec, sub2vec, graph2vec, DGI, InfoGraph.
     • Task: Graph Clustering | Datasets: MUTAG, PROTEINS, IMDB-M. | Baselines: DGI, InfoGraph.

6. Graph Representation Learning via Graphical Mutual Information Maximization Zhen Peng, Wenbing Huang, Minnan Luo, Qinghua Zheng, Yu Rong, Tingyang Xu, Junzhou Huang. WWW 2020.

   • Paper: ACM, Arxiv | Code: PyTorch
   • Method(GMI): Graphical Mutual Information
   • Experiment:
     • Task: Transductive Node Classification | Datasets: Cora, Citeseer, Pubmed. | Baselines: Raw Features, DeepWalk, EP-B, DGI, LP, Planetoid-T, GCN, GAT, GWNN.
     • Task: Inductive Node Classification | Datasets: Reddit, PPI. | Baselines: Raw Features, DeepWalk, DeepWalk+Features, GraphSAGE-GCN, GraphSAGE-Mean, GraphSAGE-LSTM, GraphSAGE-Pool, DGI, GAT, FastGCN, GaAN.
     • Task: Link Prediction | Datasets: Cora, BlogCatalog, Flickr, PPI. | Baselines: DGI.

2021

1. Bipartite Graph Embedding via Mutual Information Maximization Jiangxia Cao, Xixun Lin, Shu Guo, Luchen Liu, Tingwen Liu, Bin Wang. WSDM 2021.

   • Paper: Arxiv
   • Method(BiGI): Local-Global Mutual Information Maximization, H-hop Enclosing Subgraph
   • Experiment:
     • Task: Top-K Recommendation | Datasets: DBLP, ML-100K and ML-10M. | Baselines: DeepWalk, LINE, Node2vec, VGAE, Metapath2vec, DMGI, PinSage, BiNE, GC-MC, IGMC, NeuMF, NGCF.
     • Task: Link Prediction | Datasets: Wikipedia. | Baselines: DeepWalk, LINE, Node2vec, VGAE, Metapath2vec, DMGI, PinSage, BiNE, GC-MC, IGMC, NeuMF, NGCF.

2. How to Find Your Friendly Neighborhood: Graph Attention Design with Self-Supervision Dongkwan Kim, Alice Oh. ICLR 2021

   • Paper: OpenReview | Code: torch_geometric
   • Method(SuperGAT):
   • Experiment:
     • Task: Node Classification | Datasets: ogbn-arxiv, CS, Physics, Cora-ML, Cora-Full, DBLP, Chameleon, Four-Univ, Wiki-CS, Photo, Computers, Flickr, Crocodile, Cora, CiteSeer, PubMed, PPI. | Baselines: GCN, GraphSAGE, GAT, CGAT, GLCN, LDS, GCN + GAM, GCN + NS.
     • Task: Link Prediction | Datasets: Cora, CiteSeer, PubMed, PPI. | Baselines:

Library