MixHop: Higher-Order Graph Convolutional Architectures via Sparsified Neighborhood Mixing

30 Apr 2019Sami Abu-El-HaijaBryan PerozziAmol KapoorNazanin AlipourfardKristina LermanHrayr HarutyunyanGreg Ver SteegAram Galstyan

Existing popular methods for semi-supervised learning with Graph Neural Networks (such as the Graph Convolutional Network) provably cannot learn a general class of neighborhood mixing relationships. To address this weakness, we propose a new model, MixHop, that can learn these relationships, including difference operators, by repeatedly mixing feature representations of neighbors at various distances... (read more)

PDF Abstract

Evaluation Results from the Paper


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK COMPARE
Node Classification Citeseer MixHop Accuracy 71.4% # 19
Node Classification Citeseer MixHop Training Split 20 per node # 1
Node Classification Citeseer MixHop Validation YES # 1
Node Classification Cora MixHop Accuracy 81.9% # 24
Node Classification Cora MixHop Training Split 20 per node # 1
Node Classification Cora MixHop Validation YES # 1
Node Classification Pubmed MixHop Accuracy 80.8% # 8
Node Classification Pubmed MixHop Training Split 20 per node # 1
Node Classification Pubmed MixHop Validation YES # 1