Predict then Propagate: Graph Neural Networks meet Personalized PageRank

TL;DR

This paper introduces PPNP and APPNP, novel graph neural network models that leverage personalized PageRank to incorporate larger neighborhoods, improving semi-supervised node classification performance efficiently.

Contribution

It proposes a new propagation scheme based on personalized PageRank, enabling GNNs to utilize larger neighborhoods with better efficiency and performance.

Findings

PPNP and APPNP outperform recent GCN-like models in semi-supervised classification.

The models have comparable or faster training times and fewer parameters.

They effectively leverage large, adjustable neighborhoods for improved accuracy.

Abstract

Neural message passing algorithms for semi-supervised classification on graphs have recently achieved great success. However, for classifying a node these methods only consider nodes that are a few propagation steps away and the size of this utilized neighborhood is hard to extend. In this paper, we use the relationship between graph convolutional networks (GCN) and PageRank to derive an improved propagation scheme based on personalized PageRank. We utilize this propagation procedure to construct a simple model, personalized propagation of neural predictions (PPNP), and its fast approximation, APPNP. Our model's training time is on par or faster and its number of parameters on par or lower than previous models. It leverages a large, adjustable neighborhood for classification and can be easily combined with any neural network. We show that this model outperforms several recently proposed methods for semi-supervised classification in the most thorough study done so far for GCN-like models. Our implementation is available online.