DistShap: Scalable GNN Explanations with Distributed Shapley Values
This addresses the problem of expensive GNN explanation for users needing scalable and accurate attributions, though it is incremental as it builds on Shapley values with distributed computing.
The paper tackles the computational challenge of explaining graph neural network (GNN) predictions by proposing DistShap, a parallel algorithm that distributes Shapley value-based explanations across multiple GPUs, achieving scalability to models with millions of features using up to 128 GPUs and outperforming most existing methods in accuracy.
With the growing adoption of graph neural networks (GNNs), explaining their predictions has become increasingly important. However, attributing predictions to specific edges or features remains computationally expensive. For example, classifying a node with 100 neighbors using a 3-layer GNN may involve identifying important edges from millions of candidates contributing to the prediction. To address this challenge, we propose DistShap, a parallel algorithm that distributes Shapley value-based explanations across multiple GPUs. DistShap operates by sampling subgraphs in a distributed setting, executing GNN inference in parallel across GPUs, and solving a distributed least squares problem to compute edge importance scores. DistShap outperforms most existing GNN explanation methods in accuracy and is the first to scale to GNN models with millions of features by using up to 128 GPUs on the NERSC Perlmutter supercomputer.