How Interpretable Are Interpretable Graph Neural Networks?
This work addresses the interpretability gap in graph neural networks for scientific applications, offering a novel method to enhance subgraph extraction, though it is incremental as it builds on existing attention-based XGNNs.
The paper tackles the problem of interpretable graph neural networks (XGNNs) having limitations in accurately approximating interpretable subgraphs, leading to degenerated interpretability, and proposes a new architecture called Graph Multilinear neT (GMT) that improves interpretability and generalizability by up to 10% on benchmarks.
Interpretable graph neural networks (XGNNs ) are widely adopted in various scientific applications involving graph-structured data. Existing XGNNs predominantly adopt the attention-based mechanism to learn edge or node importance for extracting and making predictions with the interpretable subgraph. However, the representational properties and limitations of these methods remain inadequately explored. In this work, we present a theoretical framework that formulates interpretable subgraph learning with the multilinear extension of the subgraph distribution, coined as subgraph multilinear extension (SubMT). Extracting the desired interpretable subgraph requires an accurate approximation of SubMT, yet we find that the existing XGNNs can have a huge gap in fitting SubMT. Consequently, the SubMT approximation failure will lead to the degenerated interpretability of the extracted subgraphs. To mitigate the issue, we design a new XGNN architecture called Graph Multilinear neT (GMT), which is provably more powerful in approximating SubMT. We empirically validate our theoretical findings on a number of graph classification benchmarks. The results demonstrate that GMT outperforms the state-of-the-art up to 10% in terms of both interpretability and generalizability across 12 regular and geometric graph benchmarks.