GDLNN: Marriage of Programming Language and Neural Networks for Accurate and Easy-to-Explain Graph Classification
Minseok Jeon, Seunghyun Park
TL;DR
GDLNN tackles the accuracy-interpretability trade-off in graph classification by coupling a Graph Description Language (GDL) with neural networks. The GDL layer mines discriminative graph-pattern programs and encodes graphs as pattern-activation vectors, which are classified by an MLP; explanations arise from applying standard explainers to these interpretable pattern features. Empirical results on nine datasets show competitive or superior accuracy relative to GNNs and symbolic baselines, with explainability validated by qualitative and quantitative measures and a favorable cost profile when explanations are included. The approach offers a scalable, interpretable alternative to pooling-heavy graph representations and motivates further exploration of domain-specific languages in neural-symbolic graph learning.
Abstract
We present GDLNN, a new graph machine learning architecture, for graph classification tasks. GDLNN combines a domain-specific programming language, called GDL, with neural networks. The main strength of GDLNN lies in its GDL layer, which generates expressive and interpretable graph representations. Since the graph representation is interpretable, existing model explanation techniques can be directly applied to explain GDLNN's predictions. Our evaluation shows that the GDL-based representation achieves high accuracy on most graph classification benchmark datasets, outperforming dominant graph learning methods such as GNNs. Applying an existing model explanation technique also yields high-quality explanations of GDLNN's predictions. Furthermore, the cost of GDLNN is low when the explanation cost is included.