arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2203.01289cs.CVcs.LG
收藏

ADVISE：卷积神经网络的自适应特征相关性与视觉解释

ADVISE: ADaptive Feature Relevance and VISual Explanations for Convolutional Neural Networks

Universitat Oberta de Catalunya（开放加泰罗尼亚大学）

Mohammad Mahdi Dehshibi, Mona Ashtari-Majlan, Gereziher Adhane, David Masip

发表信息：
The Visual Computer, 2023, 1-13
阅读 PDF
arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文提出ADVISE方法，通过自适应带宽核密度估计量化CNN特征图单元的相关性，以提供更优的视觉解释，并在多个预训练模型上验证其优于现有方法且满足可解释性公理。

摘要：

为了使卷积神经网络（CNN）具备可解释性，必须解释不透明模型如何做出特定决策，理解导致错误的原因，改进架构设计，并识别分类器中不道德的偏见。本文介绍了ADVISE，一种新的可解释性方法，该方法量化并利用特征图中每个单元的相关性，以提供更好的视觉解释。为此，我们提出使用自适应带宽核密度估计，针对预测类别为特征图的每个单元分配相关性分数。我们还提出了一种评估协议，以定量评估CNN模型的视觉可解释性。我们在图像分类任务中使用在ImageNet上预训练的AlexNet、VGG16、ResNet50和Xception对我们的想法进行了广泛评估。我们将ADVISE与最先进的视觉可解释方法进行比较，结果表明，所提出的方法在量化特征相关性和视觉可解释性方面优于竞争方法，同时保持了有竞争力的时间复杂度。我们的实验进一步表明，ADVISE满足敏感性和实现独立性公理，同时通过了健全性检查。该实现可在https://github.com/dehshibi/ADVISE上获取，以便复现。

英文摘要：

To equip Convolutional Neural Networks (CNNs) with explainability, it is essential to interpret how opaque models take specific decisions, understand what causes the errors, improve the architecture design, and identify unethical biases in the classifiers. This paper introduces ADVISE, a new explainability method that quantifies and leverages the relevance of each unit of the feature map to provide better visual explanations. To this end, we propose using adaptive bandwidth kernel density estimation to assign a relevance score to each unit of the feature map with respect to the predicted class. We also propose an evaluation protocol to quantitatively assess the visual explainability of CNN models. We extensively evaluate our idea in the image classification task using AlexNet, VGG16, ResNet50, and Xception pretrained on ImageNet. We compare ADVISE with the state-of-the-art visual explainable methods and show that the proposed method outperforms competing approaches in quantifying feature-relevance and visual explainability while maintaining competitive time complexity. Our experiments further show that ADVISE fulfils the sensitivity and implementation independence axioms while passing the sanity checks. The implementation is accessible for reproducibility purposes on https://github.com/dehshibi/ADVISE.

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。

↑