Reliable Graph Neural Network Explanations Through Adversarial Training
This addresses the issue of unreliable explanations for GNN users, but it is incremental as it adapts an existing method from computer vision to graphs.
The paper tackles the problem of unreliable post-hoc explanations in graph neural networks by proposing adversarial training to make models more amenable to analysis, showing it helps extract domain-relevant insights in chemistry.
Graph neural network (GNN) explanations have largely been facilitated through post-hoc introspection. While this has been deemed successful, many post-hoc explanation methods have been shown to fail in capturing a model's learned representation. Due to this problem, it is worthwhile to consider how one might train a model so that it is more amenable to post-hoc analysis. Given the success of adversarial training in the computer vision domain to train models with more reliable representations, we propose a similar training paradigm for GNNs and analyze the respective impact on a model's explanations. In instances without ground truth labels, we also determine how well an explanation method is utilizing a model's learned representation through a new metric and demonstrate adversarial training can help better extract domain-relevant insights in chemistry.