A Review of Interpretability Research in Convolutional Neural Networks

A Review of Interpretability Research in Convolutional Neural Networks

by

Welcome to click the blue text to follow “Smart IT Journal“!

A Review of Interpretability Research in Convolutional Neural Networks

Dou Hui, Zhang Lingming, Han Feng, Shen Furao, Zhao Jian

Journal of Software

Journal of Software

Abstract

The performance of neural network models is increasingly powerful and widely applied to solve various computer-related tasks, demonstrating excellent capabilities. However, humans do not fully understand the operational mechanisms of neural network models. This article reviews and summarizes research on the interpretability of neural networks, discussing definitions, necessity, classifications, and evaluations of model interpretability in detail. Starting from the focus of interpretability algorithms, a new classification method for neural network interpretability algorithms is proposed, providing a fresh perspective for understanding neural networks. Based on the proposed classification method, current interpretability methods for convolutional neural networks are organized, and the characteristics of different categories of interpretability algorithms are analyzed and compared. At the same time, common evaluation principles and methods for interpretability algorithms are introduced. The research directions and applications of interpretable neural networks are overviewed, and the challenges faced by interpretable neural networks are elaborated, along with possible solutions to these challenges.

A Review of Interpretability Research in Convolutional Neural NetworksKeywords Neural Networks; Interpretability; Classification; Deep LearningA Review of Interpretability Research in Convolutional Neural Networks

Excerpt from the Paper

A Review of Interpretability Research in Convolutional Neural Networks

A Review of Interpretability Research in Convolutional Neural Networks

A Review of Interpretability Research in Convolutional Neural Networks

A Review of Interpretability Research in Convolutional Neural Networks

A Review of Interpretability Research in Convolutional Neural Networks

A Review of Interpretability Research in Convolutional Neural Networks

A Review of Interpretability Research in Convolutional Neural Networks

A Review of Interpretability Research in Convolutional Neural Networks

Welcome to visit “Journal of Software” (http://jos.org.cn) for more content. This issue’s recommended paper is

A Review of Interpretability Research in Convolutional Neural Networks

Dou Hui, Zhang Lingming, Han Feng, Shen Furao, Zhao Jian

Journal of Software, 2024, 35(1): 159-184To read the original text, please copy the link or scan the QR code belowhttp://www.jos.org.cn/jos/article/abstract/6758A Review of Interpretability Research in Convolutional Neural Networks

Copyright Statement

This article is selected from the “Journal of Software”, and the copyright belongs to “Journal of Software”. The articles reprinted by this public account are for non-profit, non-commercial educational and research purposes. We cannot contact every copyright holder for non-original articles and images used by this public account. If the authors and editors of the content believe their works should not be available online or should not be used without compensation, please notify us promptly via email or phone, and we will delete it immediately to avoid unnecessary economic losses for both parties.

A Review of Interpretability Research in Convolutional Neural Networks

Smart IT Journal

Chinese Computer Journal Network

Welcome to follow for more information

Leave a Comment