Semi-Supervised Classification with Graph Convolutional Networks

9 Sep 2016Thomas N. KipfMax Welling

We present a scalable approach for semi-supervised learning on graph-structured data that is based on an efficient variant of convolutional neural networks which operate directly on graphs. We motivate the choice of our convolutional architecture via a localized first-order approximation of spectral graph convolutions... (read more)

PDF Abstract

Evaluation results from the paper


Task Dataset Model Metric name Metric value Global rank Compare
Node Classification CiteSeer (0.5%) GCN Accuracy 43.6% # 11
Node Classification CiteSeer (1%) GCN Accuracy 55.3% # 10
Node Classification CiteSeer with Public Split: fixed 20 nodes per class GCN Accuracy 68.7% # 10
Document Classification Cora Graph-CNN Accuracy 81.5% # 5
Node Classification Cora (0.5%) GCN Accuracy 50.9% # 8
Node Classification Cora (1%) GCN Accuracy 62.3% # 8
Node Classification Cora (3%) GCN Accuracy 76.5% # 8
Node Classification Cora with Public Split: fixed 20 nodes per class GCN Accuracy 80.5% # 8
Graph Classification IPC-grounded GCN Accuracy 80.7% # 1
Graph Classification IPC-lifted GCN Accuracy 87.6% # 1
Graph Regression Lipophilicity GCN [email protected]%Train 1.05 # 2
Node Classification NELL GCN Accuracy 66.0% # 1
Node Classification PubMed (0.03%) GCN Accuracy 57.9% # 8
Node Classification PubMed (0.05%) GCN Accuracy 64.6% # 8
Node Classification PubMed (0.1%) GCN Accuracy 73.0% # 7
Node Classification PubMed with Public Split: fixed 20 nodes per class GCN Accuracy 77.8% # 9
Graph Regression Tox21 GCN [email protected]%Train 0.75 # 2