Quantized convolutional neural networks: a hardware perspective

With the rapid development of machine learning, Deep Neural Network (DNN) exhibits superior performance in solving complex problems like computer vision and natural language processing compared with classic machine learning techniques. On the other hand, the rise of the Internet of Things (IoT) and...

Full description

Saved in:
Bibliographic Details
Main Authors: Li Zhang, Olga Krestinskaya, Mohammed E. Fouda, Ahmed M. Eltawil, Khaled Nabil Salama
Format: Article
Language:English
Published: Frontiers Media S.A. 2025-07-01
Series:Frontiers in Electronics
Subjects:
Online Access:https://www.frontiersin.org/articles/10.3389/felec.2025.1469802/full
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850106708492288000
author Li Zhang
Olga Krestinskaya
Mohammed E. Fouda
Ahmed M. Eltawil
Khaled Nabil Salama
author_facet Li Zhang
Olga Krestinskaya
Mohammed E. Fouda
Ahmed M. Eltawil
Khaled Nabil Salama
author_sort Li Zhang
collection DOAJ
description With the rapid development of machine learning, Deep Neural Network (DNN) exhibits superior performance in solving complex problems like computer vision and natural language processing compared with classic machine learning techniques. On the other hand, the rise of the Internet of Things (IoT) and edge computing set a demand on executing those complex tasks on corresponding devices. As the name suggested, deep neural networks are sophisticated models with complex structures and millions of parameters, which overwhelm the capacity of IoT and edge devices. To facilitate the deployment, quantization, as one of the most promising methods, is proposed to alleviate the challenge in terms of memory usage and computation complexity by quantizing both the parameters and data flow in the DNN model into formats with shorter bit-width. Consistently, dedicated hardware accelerators are developed to further boost the execution efficiency of DNN models. In this work, we focus on Convolutional Neural Network (CNN) as an example of DNNs and conduct a comprehensive survey on various quantization and quantized training methods. We also discuss various hardware accelerator designs for quantized CNN (QCNN). Based on the review of both algorithm and hardware design, we provide general software-hardware co-design considerations. Based on the analysis, we discuss open challenges and future research directions for both algorithms and corresponding hardware designs of quantized neural networks (QNNs).
format Article
id doaj-art-40b3eb2f4c7c48138744570f2b93f301
institution OA Journals
issn 2673-5857
language English
publishDate 2025-07-01
publisher Frontiers Media S.A.
record_format Article
series Frontiers in Electronics
spelling doaj-art-40b3eb2f4c7c48138744570f2b93f3012025-08-20T02:38:46ZengFrontiers Media S.A.Frontiers in Electronics2673-58572025-07-01610.3389/felec.2025.14698021469802Quantized convolutional neural networks: a hardware perspectiveLi Zhang0Olga Krestinskaya1Mohammed E. Fouda2Ahmed M. Eltawil3Khaled Nabil Salama4Computer, Electrical and Mathematical Science and Engineering Division, King Abdullah University of Science and Technology, Thuwal, Saudi ArabiaComputer, Electrical and Mathematical Science and Engineering Division, King Abdullah University of Science and Technology, Thuwal, Saudi ArabiaRain Neuromorphics, San Francisco Inc. CA, San Francisco, United StatesComputer, Electrical and Mathematical Science and Engineering Division, King Abdullah University of Science and Technology, Thuwal, Saudi ArabiaComputer, Electrical and Mathematical Science and Engineering Division, King Abdullah University of Science and Technology, Thuwal, Saudi ArabiaWith the rapid development of machine learning, Deep Neural Network (DNN) exhibits superior performance in solving complex problems like computer vision and natural language processing compared with classic machine learning techniques. On the other hand, the rise of the Internet of Things (IoT) and edge computing set a demand on executing those complex tasks on corresponding devices. As the name suggested, deep neural networks are sophisticated models with complex structures and millions of parameters, which overwhelm the capacity of IoT and edge devices. To facilitate the deployment, quantization, as one of the most promising methods, is proposed to alleviate the challenge in terms of memory usage and computation complexity by quantizing both the parameters and data flow in the DNN model into formats with shorter bit-width. Consistently, dedicated hardware accelerators are developed to further boost the execution efficiency of DNN models. In this work, we focus on Convolutional Neural Network (CNN) as an example of DNNs and conduct a comprehensive survey on various quantization and quantized training methods. We also discuss various hardware accelerator designs for quantized CNN (QCNN). Based on the review of both algorithm and hardware design, we provide general software-hardware co-design considerations. Based on the analysis, we discuss open challenges and future research directions for both algorithms and corresponding hardware designs of quantized neural networks (QNNs).https://www.frontiersin.org/articles/10.3389/felec.2025.1469802/fullconvolutional neural networksquantizationhardwarein-memory computing (IMC)FPGA
spellingShingle Li Zhang
Olga Krestinskaya
Mohammed E. Fouda
Ahmed M. Eltawil
Khaled Nabil Salama
Quantized convolutional neural networks: a hardware perspective
Frontiers in Electronics
convolutional neural networks
quantization
hardware
in-memory computing (IMC)
FPGA
title Quantized convolutional neural networks: a hardware perspective
title_full Quantized convolutional neural networks: a hardware perspective
title_fullStr Quantized convolutional neural networks: a hardware perspective
title_full_unstemmed Quantized convolutional neural networks: a hardware perspective
title_short Quantized convolutional neural networks: a hardware perspective
title_sort quantized convolutional neural networks a hardware perspective
topic convolutional neural networks
quantization
hardware
in-memory computing (IMC)
FPGA
url https://www.frontiersin.org/articles/10.3389/felec.2025.1469802/full
work_keys_str_mv AT lizhang quantizedconvolutionalneuralnetworksahardwareperspective
AT olgakrestinskaya quantizedconvolutionalneuralnetworksahardwareperspective
AT mohammedefouda quantizedconvolutionalneuralnetworksahardwareperspective
AT ahmedmeltawil quantizedconvolutionalneuralnetworksahardwareperspective
AT khalednabilsalama quantizedconvolutionalneuralnetworksahardwareperspective