跳到主要导航 跳到搜索 跳到主要内容

"Peak" Quantization: A Training Method Suitable for Terminal Equipment to Deploy Keyword Spotting Network"

  • Xiaomeng Luo
  • , Guangcun Wei
  • , Yuhao Liu
  • , Xiaotao Jia
  • , Xinghua Yang
  • , Junlin Li
  • , Qi Wei
  • , Fei Qiao
  • Shandong University of Science and Technology
  • Beihang University
  • Beijing Forestry University
  • No.208 Research Institute of China Ordnance Industries
  • Tsinghua University

科研成果: 书/报告/会议事项章节会议稿件同行评审

摘要

In order to reduce the storage space occupied by the well-trained Keyword Spotting (KWS) network model during deployment on terminal devices while maintaining the recognition accuracy of the model as much as possible, a new method of 'peak' quantization is proposed. By limiting the maximum value of the network model weights to change the quantization process, it has achieved good results in reducing the loss of recognition accuracy. Compared with other Network quantization, such as traditional quantization and Alternating Direction Method of Multipliers (ADMM) quantization, it shows the advantages and characteristics of 'peak' quantization. Through related experiments with the Google voice dataset and the MNIST dataset, its universality is demonstrated.

源语言英语
主期刊名IProceedings of the 4th International Conference on Artificial Intelligence and Computer Engineering, ICAICE 2023
出版商Association for Computing Machinery
963-968
页数6
ISBN(电子版)9798400708831
DOI
出版状态已出版 - 17 11月 2023
活动4th International Conference on Artificial Intelligence and Computer Engineering, ICAICE 2023 - Virtual, Dalian, 中国
期限: 17 11月 202319 11月 2023

出版系列

姓名ACM International Conference Proceeding Series

会议

会议4th International Conference on Artificial Intelligence and Computer Engineering, ICAICE 2023
国家/地区中国
Virtual, Dalian
时期17/11/2319/11/23

学术指纹

探究 '"Peak" Quantization: A Training Method Suitable for Terminal Equipment to Deploy Keyword Spotting Network"' 的科研主题。它们共同构成独一无二的学术指纹。

引用此