Heterogeneous Graph Sparsification for Efficient Representation Learning

TL;DR

This paper introduces a novel sampling-based method for sparsifying heterogeneous graphs, significantly improving efficiency in representation learning while maintaining or enhancing task performance.

Contribution

It pioneers the systematic study of heterogeneous graph sparsification and develops algorithms that preserve essential information with provable sparsity guarantees.

Findings

Reduces time and space complexity in graph learning tasks.

Achieves comparable or better performance with sparser graphs.

Validated through extensive experiments on real-world data.

Abstract

Graph sparsification is a powerful tool to approximate an arbitrary graph and has been used in machine learning over homogeneous graphs. In heterogeneous graphs such as knowledge graphs, however, sparsification has not been systematically exploited to improve efficiency of learning tasks. In this work, we initiate the study on heterogeneous graph sparsification and develop sampling-based algorithms for constructing sparsifiers that are provably sparse and preserve important information in the original graphs. We have performed extensive experiments to confirm that the proposed method can improve time and space complexities of representation learning while achieving comparable, or even better performance in subsequent graph learning tasks based on the learned embedding.