Bit-serial convolution with prediction threshold for convolutional neural networks Electrical Engineering Subject Index: EL7 Signal Processing

To reduce the implementation complexity and power consumption of the convolution operation in a convolutional neural network (CNN), this work proposes a new convolution method using the serial input and prediction threshold. To confirm the benefits of the proposed method, we use the original paramet...

Full description

Saved in:
Bibliographic Details
Published inJournal of the Chinese Institute of Engineers Vol. 45; no. 3; pp. 266 - 272
Main Authors Hsiao, Jen-Hao, Chin, Wen-Long, Wu, Yu-Feng, Chang, Deng-Kai
Format Journal Article
LanguageEnglish
Published Taylor & Francis 03.04.2022
Subjects
Convolutional neural network
hardware accelerator
serial input
Online AccessGet full text

Cover

More Information
Summary:To reduce the implementation complexity and power consumption of the convolution operation in a convolutional neural network (CNN), this work proposes a new convolution method using the serial input and prediction threshold. To confirm the benefits of the proposed method, we use the original parameters, i.e. kernel weights and biases, of the popular AlexNet to verify the proposed algorithm and then implement its digital circuit. According to implementation data and comparison with traditional convolution using bit-parallel input, the implementation gain in terms of the throughput/power/area of the serial convolution is 7.57 times that of parallel convolution.
ISSN:0253-3839
2158-7299
DOI:10.1080/02533839.2022.2034050