DGI: Deep Graph Infomax¶
Deep Graph Infomax (DGI) is a general approach for learning node representations within graph-structured data in an unsupervised manner. DGI relies on maximizing mutual information between patch representations and corresponding high-level summaries of graphs—both derived using established graph convolutional network architectures.
Datasets¶
The datasets contain three citation networks: CORA, PUBMED, CITESEER. The details for these three datasets can be found in the paper.
Dependencies¶
paddlepaddle>=1.6
pgl
Performance¶
We use DGI to pretrain embeddings for each nodes. Then we fix the embedding to train a node classifier.
Dataset |
Accuracy |
---|---|
Cora |
~81% |
Pubmed |
~77.6% |
Citeseer |
~71.3% |