Semi-Supervised Classification with Graph Convolutional Networks

9 Sep 2016 Thomas N. Kipf Max 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

Results from the Paper


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK RESULT BENCHMARK
Graph Classification BP-fMRI-97 GCN Accuracy 60.7% # 4
F1 61.2% # 5
Node Classification Brazil Air-Traffic GCN (Kipf and Welling, 2017) Accuracy 0.432 # 5
Graph Classification CIFAR10 100k GCN Accuracy (%) 54.46 # 6
Node Classification Citeseer GCN Accuracy 70.3% # 39
Training Split fixed 20 per node # 3
Validation YES # 1
Node Classification CiteSeer (0.5%) GCN Accuracy 43.6% # 12
Node Classification CiteSeer (1%) GCN Accuracy 55.3% # 11
Node Classification Citeseer Full-supervised GCN Accuracy 79.34% # 3
Node Classification CiteSeer with Public Split: fixed 20 nodes per class GCN Accuracy 68.7% # 22
Document Classification Cora Graph-CNN Accuracy 81.5% # 5
Node Classification Cora GCN Accuracy 81.5% # 42
Validation YES # 1
Node Classification Cora (0.5%) GCN Accuracy 50.9% # 9
Node Classification Cora (1%) GCN Accuracy 62.3% # 9
Node Classification Cora (3%) GCN Accuracy 76.5% # 8
Node Classification Cora Full-supervised GCN Accuracy 86.64% # 4
Node Classification Cora with Public Split: fixed 20 nodes per class GCN Accuracy 80.5% # 19
Graph Classification HIV-DTI-77 GCN Accuracy 57.7% # 4
F1 54.4% # 5
Graph Classification HIV-fMRI-77 GCN Accuracy 58.3% # 5
F1 56.4& # 8
Graph Classification IPC-grounded GCN Accuracy 80.7% # 1
Graph Classification IPC-lifted GCN Accuracy 87.6% # 1
Graph Regression Lipophilicity GCN RMSE@80%Train 1.05 # 2
Node Classification NELL GCN Accuracy 66.0% # 2
Node Classification PATTERN 100k GCN Accuracy (%) 65.880 # 8
Node Classification Pubmed GCN Accuracy 79.0% # 35
Training Split fixed 20 per node # 4
Validation YES # 1
Node Classification PubMed (0.03%) GCN Accuracy 57.9% # 9
Node Classification PubMed (0.05%) GCN Accuracy 64.6% # 9
Node Classification PubMed (0.1%) GCN Accuracy 73.0% # 8
Node Classification Pubmed Full-supervised GCN Accuracy 90.22% # 4
Node Classification PubMed with Public Split: fixed 20 nodes per class GCN Accuracy 77.8% # 22
Node Classification Reddit GCN Accuracy 95.68% # 5
Skeleton Based Action Recognition SBU GCNConv Accuracy 90.00% # 9
Graph Regression Tox21 GCN AUC@80%Train 0.75 # 2
Node Classification USA Air-Traffic GCN (Kipf and Welling, 2017) Accuracy 43.2 # 6
Graph Regression ZINC 100k GCN MAE 0.469 # 8
Graph Regression ZINC-500k GCN MAE 0.367 # 12

Results from Other Papers


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK SOURCE PAPER COMPARE
Node Classification Brazil Air-Traffic GCN_cheby (Kipf and Welling, 2017) Accuracy 0.516 # 2
Node Classification Europe Air-Traffic GCN_cheby (Kipf and Welling, 2017) Accuracy 46.0 # 1
Node Classification Europe Air-Traffic GCN (Kipf and Welling, 2017) Accuracy 37.1 # 5
Node Classification Facebook GCN_cheby (Kipf and Welling, 2017) Accuracy 64.6 # 2
Node Classification Facebook GCN (Kipf and Welling, 2017) Accuracy 57.5 # 4
Node Classification Flickr GCN (Kipf and Welling, 2017) Accuracy 0.546 # 4
Node Classification Flickr GCN_cheby (Kipf and Welling, 2017) Accuracy 0.479 # 5
Node Classification Wiki-Vote GCN_cheby (Kipf and Welling, 2017) Accuracy 49.5 # 3
Node Classification Wiki-Vote GCN (Kipf and Welling, 2017) Accuracy 32.9 # 5

Methods used in the Paper


METHOD TYPE
GCN
Graph Models