GANMEX: One-vs-One Attributions using GAN-based Model Explainability
Sheng-Min Shih, Pin-Ju Tien, Zohar S. Karnin
Abstract
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 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. Achieving one-vs-one 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 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 baseline as the closest realistic sample belong to the target class, which allows attribution methods to provide true one-vs-one 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.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
2 top-tier papers have cited it since. Know what they found before your reviewers do.
Lune can draw on
the full paper · the top-tier papers that cite it · related work
Free to start. No credit card required.
Lineage
- StarGAN v2: Diverse Image Synthesis for Multiple Domains cites Few-Shot Unsupervised Image-to-Image Translation
- Rethinking Visual Counterfactual Explanations Through Region Constraint cites Adversarial Counterfactual Visual Explanations
Cited by top-tier papers2
Builds on4
- Few-Shot Unsupervised Image-to-Image TranslationMing-Yu Liu, Xun Huang, Arun Mallya, Tero Karras et al.ICCV 2019 · 668 citations
- Sanity Checks for Saliency MetricsRichard Tomsett, Dan Harborne, Supriyo Chakraborty, Prudhvi Gurram et al.AAAI 2020 · 204 citations
- Concise Explanations of Neural Networks using Adversarial TrainingPrasad Chalasani, Jiefeng Chen, Amrita Roy Chowdhury, Xi Wu et al.ICML 2020 · 148 citations
- StarGAN v2: Diverse Image Synthesis for Multiple DomainsYunjey Choi, Youngjung Uh, Jaejun Yoo, Jung-Woo HaCVPR 2020
Related papers
- Explaining in Style: Training a GAN to explain a classifier in StyleSpaceOran Lang, Yossi Gandelsman, Michal Yarom, Yoav Wald et al.ICCV 2021 · 181 citations
- Building Reliable Explanations of Unreliable Neural Networks: Locally Smoothing Perspective of Model InterpretationDohun Lim, Hyeonseok Lee, Sungchan KimCVPR 2021
- Attribute Manipulation Generative Adversarial Networks for Fashion ImagesKenan E. Ak, Ashraf A. Kassim, Joo-Hwee Lim, Jo Yew ThamICCV 2019 · 85 citations
- An Adversarial Framework for Generating Unseen Images by Activation MaximizationYang Zhang, Wang Zhou, Gaoyuan Zhang, David D. Cox et al.AAAI 2022 · 1 citation
- ReMix: Towards Image-to-Image Translation With Limited DataJie Cao, Luanxuan Hou, Ming-Hsuan Yang, Ran He et al.CVPR 2021
