GANMEX: One-vs-One Attributions Guided by GAN-based Counterfactual Explanation Baselines
This addresses the issue of explaining class differences in multi-class classifiers, particularly for similar classes, but is incremental as it builds on existing attribution methods with a novel baseline selection approach.
The paper tackles the problem of baseline selection in attribution methods for multi-class classifiers, which limits one-vs-one explanations, and shows that GANMEX improves saliency maps and outperforms existing baselines on perturbation-based metrics and model randomization tests.
Attribution methods have been shown as promising approaches for identifying key features that led to learned model predictions. While most existing attribution methods rely on a baseline input for performing feature perturbations, limited research has been conducted to address the baseline selection issues. Poor choices of baselines limit the ability of one-vs-one (1-vs-1) explanations for multi-class classifiers, which means the attribution methods were not able to explain why an input belongs to its original class but not the other specified target class. 1-vs-1 explanation is crucial when certain classes are more similar than others, e.g. two bird types among multiple animals, by focusing on key differentiating features rather than shared features across classes. In this paper, we present GAN-based Model EXplainability (GANMEX), a novel approach applying Generative Adversarial Networks (GAN) by incorporating the to-be-explained classifier as part of the adversarial networks. Our approach effectively selects the counterfactual baseline as the closest realistic sample belong to the target class, which allows attribution methods to provide true 1-vs-1 explanations. We showed that GANMEX baselines improved the saliency maps and led to stronger performance on perturbation-based evaluation metrics over the existing baselines. Existing attribution results are known for being insensitive to model randomization, and we demonstrated that GANMEX baselines led to better outcome under the cascading randomization of the model.