Scalable Graph Neural Networks for Heterogeneous Graphs

19 Nov 2020  ·  Lingfan Yu, Jiajun Shen, Jinyang Li, Adam Lerer ·

Graph neural networks (GNNs) are a popular class of parametric model for learning over graph-structured data. Recent work has argued that GNNs primarily use the graph for feature smoothing, and have shown competitive results on benchmark tasks by simply operating on graph-smoothed node features, rather than using end-to-end learned feature hierarchies that are challenging to scale to large graphs. In this work, we ask whether these results can be extended to heterogeneous graphs, which encode multiple types of relationship between different entities. We propose Neighbor Averaging over Relation Subgraphs (NARS), which trains a classifier on neighbor-averaged features for randomly-sampled subgraphs of the "metagraph" of relations. We describe optimizations to allow these sets of node features to be computed in a memory-efficient way, both at training and inference time. NARS achieves a new state of the art accuracy on several benchmark datasets, outperforming more expensive GNN-based methods

PDF Abstract

Datasets


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Node Property Prediction ogbn-mag NARS Test Accuracy 0.5240 ± 0.0016 # 20
Validation Accuracy 0.5372 ± 0.0009 # 20
Number of params 4130149 # 32
Ext. data No # 1

Methods


No methods listed for this paper. Add relevant methods here