Please use this identifier to cite or link to this item:
http://hdl.handle.net/1942/37913
Title: | Sample-Wise Dynamic Precision Quantization for Neural Network Acceleration | Authors: | LI, Bowen Xiong, Dongliang Huang, Kai Jiang, Xiaowen Yao, Hao Chen , Junjian CLAESEN, Luc |
Issue Date: | 2022 | Publisher: | IEICE-INST ELECTRONICS INFORMATION COMMUNICATION ENGINEERS | Source: | IEICE Electronics Express, 19 (16), p. 1-6 | Abstract: | Quantization is a well-known method for deep neural networks (DNNs) compression and acceleration. In this work, we propose the Sample-Wise Dynamic Precision (SWDP) quantization scheme, which can switch the bit-width of weights and activations in the model according to the task difficulty of input samples at runtime. Using low-precision networks for easy input images brings advantages in terms of computational and energy efficiency. We also propose an adaptive hardware design for the efficient implementation of our SWDP networks. The experimental results on various networks and datasets demonstrate that our SWDP achieves an average of 3.3x speedup and 3.0x energy saving over the bit-level dynamically composable architecture BitFusion. | Notes: | Xiong, DL (corresponding author), Zhejiang Univ, Sch Micronano Elect, Hangzhou 310030, Peoples R China. xiongdl@zju.edu.cn |
Keywords: | convolutional neural networks;dynamic quantization;hardware accelerators | Document URI: | http://hdl.handle.net/1942/37913 | ISSN: | 1349-2543 | e-ISSN: | 1349-2543 | DOI: | 10.1587/elex.19.20220229 | ISI #: | 000826303400001 | Rights: | © 2019 The Institute of Electronics, Information and Communication Engineer | Category: | A1 | Type: | Journal Contribution | Validations: | ecoom 2023 |
Appears in Collections: | Research publications |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
19_19.20220229.pdf | Published version | 2.68 MB | Adobe PDF | View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.