Research map / Label-efficient and robust learning / Recognition and long-tail
ExplainGAN: Model Explanation via Decision Boundary Crossing Transformations
Pouya Samangouei, Ardavan Saeedi, Liam Nakagawa, Nathan Silberman
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add reconstruction loss/path (ablation) | ExplainGAN · Medical Ultrasound substitutability (A2C vs A4C), Both Transformed | accuracy (%): 96.1 -> 97.3 | Improved |
| add reconstruction loss/path (ablation) | ExplainGAN · Medical Ultrasound substitutability (A2C vs A4C), Transformed 1 | accuracy (%): 83.9 -> 94.2 | Improved |
| add mask output (ablation) | ExplainGAN · Medical Ultrasound substitutability (A2C vs A4C), Both Transformed | accuracy (%): 95.1 -> 97.3 | Improved |
| add mask output (ablation) | ExplainGAN · Medical Ultrasound substitutability (A2C vs A4C), Transformed 1 | accuracy (%): 97.3 -> 94.2 | Hurt |
| add mask output (ablation) | ExplainGAN · Medical Ultrasound substitutability (A2C vs A4C), Transformed 0 | accuracy (%): 93.9 -> 95.5 | Improved |
| use ExplainGAN instead of CycleGAN | ExplainGAN vs CycleGAN · Medical Ultrasound substitutability (A2C vs A4C), Both Transformed | accuracy (%): 84.1 -> 97.3 | Improved |
| use ExplainGAN instead of CycleGAN | ExplainGAN vs CycleGAN · Medical Ultrasound substitutability (A2C vs A4C), Transformed 1 | accuracy (%): 64 -> 94.2 | Improved |
| use ExplainGAN masks instead of Grad saliency | ExplainGAN vs Grad · MNIST AOSPC (threshold 0.5) | AOSPC: 1474 -> 2622 | Improved |
| use ExplainGAN masks instead of Occlusion at high threshold | ExplainGAN vs Occlusion · MNIST AOSPC (threshold 0.9) | AOSPC: 1486 vs 1474 | No clear effect |
| use ExplainGAN masks instead of Occlusion | ExplainGAN vs Occlusion · Ultrasound AOSPC (threshold 0.5) | AOSPC: 1215 vs 1167 | Hurt |
| use ExplainGAN masks instead of Occlusion at low threshold | ExplainGAN vs Occlusion · Ultrasound AOSPC (threshold 0.9) | AOSPC: 142 -> 374 | Improved |
| transformed images used for training on CelebA mustache | ExplainGAN · CelebA Mustache substitutability | accuracy (%): 76.65 vs 83.9 | Hurt |
Closest papers by meaning
- Explaining in Diffusion: Explaining a Classifier with Diffusion Semantics (CVPR 2025 · similarity 0.87)
- Language Model as Visual Explainer (NeurIPS 2024 · similarity 0.87)
- Probabilistic Conceptual Explainers: Trustworthy Conceptual Explanations for Vision Foundation Models (ICML 2024 · similarity 0.87)
- Comparing the Decision-Making Mechanisms by Transformers and CNNs via Explanation Methods (CVPR 2024 · similarity 0.87)
- Restyling Unsupervised Concept Based Interpretable Networks with Generative Models (ICLR 2025 · similarity 0.86)
- Explainable Concept Generation through Vision-Language Preference Learning for Understanding Neural Networks' Internal Representations (ICML 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
