arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2303.07580cs.LGcs.CV
收藏

基于可解释AI的敏感区域变形测试框架

Sensitive Region-based Metamorphic Testing Framework using Explainable AI

AGEST, Inc.(AGEST公司)

Yuma Torikoshi, Yasuharu Nishi, Juichi Takahashi

摘要:

深度学习是机器学习中最热门的研究课题之一,由深度学习驱动的图像识别系统发展迅速。最近的研究采用变形测试(MT)来检测错误分类的图像。他们大多数讨论变质关系(MR),与有限的关注给予哪些地区应该被改造。我们专注于这样一个事实,即有敏感区域,即使是小的变换也可以很容易地改变预测结果,并提出了一个MT框架,通过变换这些敏感区域来有效地测试容易发生错误分类的区域。我们的评估表明,敏感区域可以由可解释AI(XAI)指定,并且我们的框架可以有效地检测故障。

英文摘要:

Deep Learning (DL) is one of the most popular research topics in machine learning and DL-driven image recognition systems have developed rapidly. Recent research has employed metamorphic testing (MT) to detect misclassified images. Most of them discuss metamorphic relations (MR), with limited attention given to which regions should be transformed. We focus on the fact that there are sensitive regions where even small transformations can easily change the prediction results and propose an MT framework that efficiently tests for regions prone to misclassification by transforming these sensitive regions. Our evaluation demonstrated that the sensitive regions can be specified by Explainable AI (XAI) and our framework effectively detects faults.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑