Scalable Graph Neural Networks for Heterogeneous Graphs
This work provides a more scalable and memory-efficient approach for applying GNNs to large heterogeneous graphs, which is a problem for researchers and practitioners dealing with complex, multi-relational data.
This paper investigates whether the success of graph-smoothed node features in GNNs can be extended to heterogeneous graphs. The authors propose Neighbor Averaging over Relation Subgraphs (NARS), which trains a classifier on neighbor-averaged features from randomly sampled relation subgraphs, achieving new state-of-the-art accuracy on several benchmark datasets.
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