Revisiting Semi-Supervised Learning with Graph Embeddings

29 Mar 2016Zhilin Yang • William W. Cohen • Ruslan Salakhutdinov

We present a semi-supervised learning framework based on graph embeddings. Given a graph between instances, we train an embedding for each instance to jointly predict the class label and the neighborhood context in the graph. We develop both transductive and inductive variants of our method.

Full paper

Evaluation


Task Dataset Model Metric name Metric value Global rank Compare
Node Classification Citeseer Planetoid* Accuracy 64.7% # 8
Node Classification Cora Planetoid* Accuracy 75.7% # 8
Node Classification NELL Planetoid* Accuracy 61.9% # 2
Node Classification Pubmed Planetoid* Accuracy 77.2% # 6