Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering

NeurIPS 2016 Michaël DefferrardXavier BressonPierre Vandergheynst

In this work, we are interested in generalizing convolutional neural networks (CNNs) from low-dimensional regular grids, where image, video and speech are represented, to high-dimensional irregular domains, such as social networks, brain connectomes or words' embedding, represented by graphs. We present a formulation of CNNs in the context of spectral graph theory, which provides the necessary mathematical background and efficient numerical schemes to design fast localized convolutional filters on graphs... (read more)

PDF Abstract
Task Dataset Model Metric name Metric value Global rank Compare
Node Classification Citeseer ChebNet Accuracy 69.8% # 18
Node Classification CiteSeer (0.5%) ChebyNet Accuracy 45.3% # 8
Node Classification CiteSeer (1%) ChebyNet Accuracy 59.4% # 8
Node Classification CiteSeer with Public Split: fixed 20 nodes per class ChebyNet Accuracy 70.1% # 8
Node Classification Cora ChebNet Accuracy 81.2% # 20
Node Classification Cora (0.5%) ChebyNet Accuracy 33.9% # 14
Node Classification Cora (1%) ChebyNet Accuracy 44.2% # 14
Node Classification Cora (3%) ChebyNet Accuracy 62.1% # 13
Node Classification Cora with Public Split: fixed 20 nodes per class ChebyNet Accuracy 78.0% # 12
Node Classification Pubmed ChebNet Accuracy 74.4% # 18
Node Classification PubMed (0.03%) ChebyNet Accuracy 45.3% # 14
Node Classification PubMed (0.05%) ChebyNet Accuracy 48.2% # 14
Node Classification PubMed (0.1%) ChebyNet Accuracy 55.2% # 14
Node Classification PubMed with Public Split: fixed 20 nodes per class ChebyNet Accuracy 69.8% # 14