Towards Causal Classification: A Comprehensive Study on Graph Neural Networks
Simi Job, Xiaohui Tao, Taotao Cai, Lin Li, Haoran Xie, Jianming Yong
TL;DR
This study examines whether incorporating causal reasoning into Graph Neural Networks (GNNs) improves graph classification. It benchmarks nine architectures (including causal and mutual-information–based variants) across seven diverse datasets to assess generalizability and the true value of causality in GNNs. Key findings show that while attention-based causal models like CAL can be competitive, they do not consistently outperform strong baselines, and mutual-information–based approaches often underperform, highlighting the need for more robust causal mechanisms. The work underscores the importance of hyperparameter tuning and cross-domain evaluation to gauge when and how causal GNNs can offer real benefits for graph-structured data.
Abstract
The exploration of Graph Neural Networks (GNNs) for processing graph-structured data has expanded, particularly their potential for causal analysis due to their universal approximation capabilities. Anticipated to significantly enhance common graph-based tasks such as classification and prediction, the development of a causally enhanced GNN framework is yet to be thoroughly investigated. Addressing this shortfall, our study delves into nine benchmark graph classification models, testing their strength and versatility across seven datasets spanning three varied domains to discern the impact of causality on the predictive prowess of GNNs. This research offers a detailed assessment of these models, shedding light on their efficiency, and flexibility in different data environments, and highlighting areas needing advancement. Our findings are instrumental in furthering the understanding and practical application of GNNs in diverse datacentric fields