Treffer: An accurate pixel-Level explainable approach for CNNs and its application.

Title:
An accurate pixel-Level explainable approach for CNNs and its application.
Authors:
Zhang H; School of Information Science & Engineering, Lanzhou University, No. 222, Tianshui South Road, Chengguan District, Lanzhou, 730000, Gansu, China., Wang J; School of Information Science & Engineering, Lanzhou University, No. 222, Tianshui South Road, Chengguan District, Lanzhou, 730000, Gansu, China., Wang Z; School of Information Science & Engineering, Lanzhou University, No. 222, Tianshui South Road, Chengguan District, Lanzhou, 730000, Gansu, China., Li G; Shanghai Jiao Tong University, No. 800, Dongchuan Road, Minhang District, Shanghai, 200240, China., Cheng Z; Jiangxi Provincial Key Laboratory for High Performance Computing, State International Science & Technology Cooperation Base of Networked Supporting Software, School of Artificial Intelligence, Jiangxi Normal University, No. 99, Ziyang Avenue, Nanchang, 330000, Jiangxi, China. Electronic address: zhuo_cheng@126.com.
Source:
Neural networks : the official journal of the International Neural Network Society [Neural Netw] 2026 Mar; Vol. 195, pp. 108289. Date of Electronic Publication: 2025 Nov 04.
Publication Type:
Journal Article
Language:
English
Journal Info:
Publisher: Pergamon Press Country of Publication: United States NLM ID: 8805018 Publication Model: Print-Electronic Cited Medium: Internet ISSN: 1879-2782 (Electronic) Linking ISSN: 08936080 NLM ISO Abbreviation: Neural Netw Subsets: MEDLINE
Imprint Name(s):
Original Publication: New York : Pergamon Press, [c1988-
Contributed Indexing:
Keywords: Artificial intelligence; CNN; Explainable AI; Image processing; Robustness; Symbolic execution
Entry Date(s):
Date Created: 20251113 Date Completed: 20260124 Latest Revision: 20260128
Update Code:
20260130
DOI:
10.1016/j.neunet.2025.108289
PMID:
41232222
Database:
MEDLINE

Weitere Informationen

Convolutional neural network (CNN) has been widely used to undertake the task of image classification. Unfortunately, the implicit decision knowledge carried by a trained CNN model is often difficult to be comprehended by humans. A feasible method to make human understanding of decision knowledge is to interpret the classification behaviour of the trained CNN model. Currently, many explainable methods have been proposed and made great contributions. However, the existing methods often suffer from insufficient interpretation accuracy. This paper presents a novel pixel-level explainable approach to address the problem of lack of interpretation accuracy in the existing methods. A large number of experiments are conducted on the PyTorch team published CNN models, and the experimental results show that the presented approach is a 100 % accurate technique for interpreting classification basis of input images on pixel-level in comparison to the existing explainable methods. In addition, a scheme to enhance the adversarial robustness of CNN models is designed based on the presented explainable approach. The evaluation experiments show that the designed scheme provides an effective way to improve the adversarial robustness of the CNN models, and is a transferable technique for the different structure CNN models.
(Copyright © 2025 Elsevier Ltd. All rights reserved.)

Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.