Guided AbsoluteGrad: Magnitude of Gradients Matters to Explanation's Localization and Saliency
This work addresses the need for more localized and less noisy explanations in XAI for domains like image classification, though it is incremental as it builds on existing gradient-based approaches.
The paper tackled the problem of improving saliency map explanations in gradient-based XAI methods by proposing Guided AbsoluteGrad, which uses gradient magnitudes and variance to reduce noise, and it outperformed seven other methods in evaluations on datasets like ImageNet, ISIC, and Places365.
This paper proposes a new gradient-based XAI method called Guided AbsoluteGrad for saliency map explanations. We utilize both positive and negative gradient magnitudes and employ gradient variance to distinguish the important areas for noise deduction. We also introduce a novel evaluation metric named ReCover And Predict (RCAP), which considers the Localization and Visual Noise Level objectives of the explanations. We propose two propositions for these two objectives and prove the necessity of evaluating them. We evaluate Guided AbsoluteGrad with seven gradient-based XAI methods using the RCAP metric and other SOTA metrics in three case studies: (1) ImageNet dataset with ResNet50 model; (2) International Skin Imaging Collaboration (ISIC) dataset with EfficientNet model; (3) the Places365 dataset with DenseNet161 model. Our method surpasses other gradient-based approaches, showcasing the quality of enhanced saliency map explanations through gradient magnitude.