TopInG: Topologically Interpretable Graph Learning via Persistent Rationale Filtration

Graph Neural Networks (GNNs) have shown remarkable success across various scientific fields,yet their adoption in critical decision-making is often hindered by a lack of interpretability. Recently,intrinsic interpretable GNNs have been studied to provide insights into model predictions by identifying rationale substructures in graphs. However, existing methods face challenges when the underlying rationale subgraphs are complex and varied. In this work, we propose TopInG: Topologically Interpretable Graph Learning, a novel topological framework that leverages persistent homology to identify persistent rationale subgraphs. TopInG employs a rationale filtration learning approach to model an autoregressive generating process of rationale subgraphs, and introduces a self-adjusted topological constraint, termed topological discrepancy, to enforce a persistent topological distinction between rationale subgraphs and irrelevant counterparts. We provide theoretical guarantees that our loss function is uniquely optimized by the ground truth under specific conditions. Extensive experiments demonstrate TopInG's effectiveness in tackling key challenges, such as handling variform rationale subgraphs, balancing predictive performance with interpretability, and mitigating spurious correlations. Results show that our approach improves upon state-of-the-artmethods on both predictive accuracy and interpretation quality.

Powerful AI models known as Graph Neural Networks (GNNs) are increasingly used in science, but their "black box" nature can make them difficult to trust for critical decisions. A key challenge is explaining why a GNN makes a certain prediction, especially when the underlying reasons can have many different shapes and structures. For instance, in biology, different molecules might cause the same effect through entirely different structural components. Our work, TopInG, addresses this by teaching the AI to focus on the fundamental shape and structure of the data using a mathematical field called topology. The method learns to identify and prioritize the most important parts of a data structure, separating them from less relevant components. This process creates a clear structural gap, allowing the model to reliably distinguish between the essential "rationale" for a decision and the background noise.This approach leads to more trustworthy and transparent AI, allowing scientists to understand the "why" behind a prediction, not just the "what". Our experiments show that TopInG is more effective than previous methods at identifying the correct explanation, especially when dealing with complex and diverse data. This helps build more reliable AI tools for scientific discovery.