On Calibration of Graph Neural Networks for Node Classification
This addresses the need for reliable confidence estimates in safety-critical applications using graph neural networks, representing an incremental advancement in calibration techniques.
The paper tackles the problem of miscalibrated confidence scores in graph neural networks for node classification, proposing a topology-aware calibration method that improves calibration compared to baseline methods, with specific numerical gains mentioned in the abstract.
Graphs can model real-world, complex systems by representing entities and their interactions in terms of nodes and edges. To better exploit the graph structure, graph neural networks have been developed, which learn entity and edge embeddings for tasks such as node classification and link prediction. These models achieve good performance with respect to accuracy, but the confidence scores associated with the predictions might not be calibrated. That means that the scores might not reflect the ground-truth probabilities of the predicted events, which would be especially important for safety-critical applications. Even though graph neural networks are used for a wide range of tasks, the calibration thereof has not been sufficiently explored yet. We investigate the calibration of graph neural networks for node classification, study the effect of existing post-processing calibration methods, and analyze the influence of model capacity, graph density, and a new loss function on calibration. Further, we propose a topology-aware calibration method that takes the neighboring nodes into account and yields improved calibration compared to baseline methods.