Multi-Task Graph Autoencoders

7 Nov 2018Phi Vu Tran

We examine two fundamental tasks associated with graph representation learning: link prediction and node classification. We present a new autoencoder architecture capable of learning a joint representation of local graph structure and available node features for the simultaneous multi-task learning of unsupervised link prediction and semi-supervised node classification. Our simple, yet effective and versatile model is efficiently trained end-to-end in a single stage, whereas previous related deep graph embedding methods require multiple training steps that are difficult to optimize.

Full paper

Evaluation


Task Dataset Model Metric name Metric value Global rank Compare
Node Classification Citeseer MTGAE Accuracy 71.80% # 4
Link Prediction Citeseer MTGAE Accuracy 94.90% # 1
Node Classification Cora MTGAE Accuracy 79.00% # 6
Link Prediction Cora MTGAE Accuracy 94.60% # 1
Node Classification Pubmed MTGAE Accuracy 80.40% # 1
Link Prediction Pubmed MTGAE Accuracy 94.40% # 2