Explaining image classifiers by removing input features using generative models
This work addresses the issue of generating more realistic explanations for image classifiers, which is important for interpretability in AI, though it is incremental as it builds on existing attribution methods.
The paper tackled the problem of unrealistic counterfactual samples in perturbation-based explanation methods for image classifiers by integrating a generative inpainter, resulting in improved plausibility, accuracy across three metrics, and robustness across datasets and model pairs.
Perturbation-based explanation methods often measure the contribution of an input feature to an image classifier's outputs by heuristically removing it via e.g. blurring, adding noise, or graying out, which often produce unrealistic, out-of-samples. Instead, we propose to integrate a generative inpainter into three representative attribution methods to remove an input feature. Our proposed change improved all three methods in (1) generating more plausible counterfactual samples under the true data distribution; (2) being more accurate according to three metrics: object localization, deletion, and saliency metrics; and (3) being more robust to hyperparameter changes. Our findings were consistent across both ImageNet and Places365 datasets and two different pairs of classifiers and inpainters.