面向可解释深度神经网络的特征联合扰动
计算机视觉与模式识别
2024-08-27 v1 机器学习
摘要
深度神经网络（DNN）固有的“黑箱”特性损害了其透明度和可靠性。最近，解释性人工智能（XAI）受到研究者的越来越大关注。several perturbation-based interpretations have emerged。然而，这些方法往往不充分考虑特征之间的依赖关系。为解决此问题，我们提出了一种由特征联合驱动的扰动解释方法，该方法利用网络的深层信息提取相关特征。然后，我们设计了 carefully-designed consistency loss 来指导网络解释。conduct both quantitative and qualitative experiments 以验证我们所提出方法的有效性。Code is available at github.com/Teriri1999/Perturebation-on-Feature-Coalition.。
引用
@article{arxiv.2408.13397,
title = {Perturbation on Feature Coalition: Towards Interpretable Deep Neural Networks},
author = {Xuran Hu and Mingzhe Zhu and Zhenpeng Feng and Miloš Daković and Ljubiša Stanković},
journal= {arXiv preprint arXiv:2408.13397},
year = {2024}
}
备注
4 pages, 4 figures, 2 tables