arXiv：cs.LG· Josef Lindl, Mariana Chaves, Damien Garreau·· 3 小时前
用生成式图像修复改进 LIME：让图像 XAI 扰动更真实
Inpainting Insights: Elevating Visual XAI with Photorealistic Perturbations
AI 导读
研究者调整了基于扰动的解释方法 LIME，用生成式图像修复替代替换像素值的传统扰动方式，生成与原始数据分布更接近的逼真扰动样本，从而提升图像解释质量。原有方法（包括更精细的确定性技术）会产生不真实的分布外样本并留下可见伪影，可能误导模型。
正文
Abstract:The increasing complexity of state-of-the-art machine learning models has made their behavior progressively harder to interpret, spurring rapid advancements in the field of eXplainable Artificial Intelligence (XAI). Among many methods proposed, perturbation-based approaches play a major role. By systematically altering (perturbing) input features, these approaches measure the impact on the model's predictions. For image data, traditional perturbation techniques, often involve replacing pixel values e.g., with a pre-defined color. However, such approaches, but also more refined deterministic techniques, generate unrealistic out-of-distribution samples and often leave visible artifacts, which can mislead the model and compromise explanation quality. In this work, we adjust LIME, a widely used perturbation-based method, to demonstrate how generative inpainting can improve perturbation-based explanations for images. We achieve photorealistic perturbed samples that align better with the original data distribution and enhance explanation quality.
| Comments: | Accepted preprint (v2, updated affiliations). 15 pages, 5 figures, 4 tables. This preprint has not undergone peer review or any post-submission improvements or corrections. The Version of Record of this contribution is published in ICPR 2026 International Workshops. ICPR 2026. Lecture Notes in Computer Science, vol 17114, and is available online at this https URL |
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2607.15482 [cs.LG] |
| (or arXiv:2607.15482v2 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2607.15482 arXiv-issued DOI via DataCite |
|
| Journal reference: | Pattern Recognition. ICPR 2026 International Workshops. ICPR 2026. Lecture Notes in Computer Science, vol 17114. Springer, Cham, 2026, pp. 447-461 |
| Related DOI: | https://doi.org/10.1007/978-3-032-39388-3_32
DOI(s) linking to related resources |
Submission history
From: Josef Lindl [view email]
[v1]
Thu, 16 Jul 2026 21:57:27 UTC (15,828 KB)
[v2]
Wed, 14 Oct 2026 09:30:54 UTC (15,828 KB)
来源：arXiv：cs.LG · arxiv.org