Evaluating the Use of Interpretable Quantized Convolutional Neural Networks for Resource-Constrained Deployment

Tools

Rogers, Harry ORCID: https://orcid.org/0000-0003-3227-5677, De La Iglesia, Beatriz ORCID: https://orcid.org/0000-0003-2675-5826 and Zebin, Tahmina ORCID: https://orcid.org/0000-0003-0437-0570 (2023) Evaluating the Use of Interpretable Quantized Convolutional Neural Networks for Resource-Constrained Deployment. In: Proceedings of the 15th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management. International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management, IC3K - Proceedings, 1 . SciTePress – Science and Technology Publications, pp. 109-120. ISBN 978-989-758-671-2

Preview

PDF (122319) - Published Version
Available under License Creative Commons Attribution Non-commercial No Derivatives.
Download (35MB) | Preview

Abstract

The deployment of Neural Networks on resource-constrained devices for object classification and detection has led to the adoption of network compression methods, such as Quantization. However, the interpretation and comparison of Quantized Neural Networks with their Non-Quantized counterparts remains inadequately explored. To bridge this gap, we propose a novel Quantization Aware eXplainable Artificial Intelligence (XAI) pipeline to effectively compare Quantized and Non-Quantized Convolutional Neural Networks (CNNs). Our pipeline leverages Class Activation Maps (CAMs) to identify differences in activation patterns between Quantized and Non-Quantized. Through the application of Root Mean Squared Error, a subset from the top 5% scoring Quantized and Non-Quantized CAMs is generated, highlighting regions of dissimilarity for further analysis. We conduct a comprehensive comparison of activations from both Quantized and Non-Quantized CNNs, using Entropy, Standard Deviation, Sparsity metric s, and activation histograms. The ImageNet dataset is utilized for network evaluation, with CAM effectiveness assessed through Deletion, Insertion, and Weakly Supervised Object Localization (WSOL). Our findings demonstrate that Quantized CNNs exhibit higher performance in WSOL and show promising potential for real-time deployment on resource-constrained devices.

Item Type:	Book Section
Additional Information:	Funding Information: This work is supported by the Engineering and Physical Sciences Research Council [EP/S023917/1]. Publisher Copyright: Copyright © 2023 by SCITEPRESS - Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)
Uncontrolled Keywords:	class activation maps,deep learning,quantization,xai,software,management of technology and innovation,strategy and management ,/dk/atira/pure/subjectarea/asjc/1700/1712
Faculty \ School:	Faculty of Science Faculty of Science > School of Computing Sciences
UEA Research Groups:	Faculty of Science > Research Groups > Norwich Epidemiology Centre Faculty of Medicine and Health Sciences > Research Groups > Norwich Epidemiology Centre Faculty of Medicine and Health Sciences > Research Centres > Norwich Institute for Healthy Aging Faculty of Science > Research Groups > Data Science and Statistics
Related URLs:	http://www.scopus.com/inward/record.url?...
Depositing User:	LivePure Connector
Date Deposited:	22 Dec 2023 02:28
Last Modified:	24 Apr 2024 14:30
URI:	https://ueaeprints.uea.ac.uk/id/eprint/94040
DOI:	10.5220/0012231900003598

Downloads

Downloads per month over past year

Actions (login required)

View Item