SynHING: Synthetic Heterogeneous Information Network Generation for Graph Learning and Explanation

Ming-Yi Hong, Yi-Hsiang Huang, Shao-En Lin, You-Chen Teng, Chih-Yu Wang, Che Lin

Keywords: synthetic graph generation, heterogeneous information networks, graph neural networks, explainable artificial intelligence

Abstract:

Graph Neural Networks (GNNs) excel in modeling graph structures across diverse domains, such as community analysis and recommendation systems. As the need for GNN interpretability grows, there is an increasing demand for robust baselines and comprehensive graph datasets, especially within the realm of Heterogeneous Information Networks (HIN). To address this, we introduce SynHING, a framework for Synthetic Heterogeneous Information Network Generation designed to advance graph learning and explanation.
After identifying key motifs in a target HIN, SynHING systematically employs a bottom-up generation process with intra-cluster and inter-cluster merge modules. This process, along with post-pruning techniques, ensures that the synthetic HIN accurately mirrors the structural and statistical properties of the original graph. The effectiveness of SynHING is validated using four datasets - IMDB, Recipe, ACM, and DBLP - spanning three distinct application categories, demonstrating both its generality and practicality.
Furthermore, SynHING provides ground-truth motifs for evaluating GNN explainer models, establishing a new benchmark for explainable, synthetic HIN generation. This contributes significantly to advancing interpretable machine learning in complex network environments.

Primary Area: learning on graphs and other geometries & topologies

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 8484