Energy efficient CNN inference accelerator using fast fourier transform

Ya Chin Chung, Po Hsiang Cheng, Chih Wei Liu

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

We use FFT-based convolution in frequency domain to reduce computational complexity in CNNs. The properties of conjugate symmetry and down-sampling is adopted to further reduce complexity. By eliminating filter weights in CNNs that can save computational requirement but lead to accuracy loss. The simulation result reveals that eliminating filter weights in frequency domain is more accurate than that in time domain. With the proposed design synthesized by TSMC 90 nm CMOS technology, the total latency, power and energy are considerably competitive. As a result, our FFT-based CNN accelerator is energy-efficient.

Original languageEnglish
Title of host publication2019 International Symposium on VLSI Design, Automation and Test, VLSI-DAT 2019
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781728106557
DOIs
StatePublished - Apr 2019
Event2019 International Symposium on VLSI Design, Automation and Test, VLSI-DAT 2019 - Hsinchu, Taiwan
Duration: 22 Apr 201925 Apr 2019

Publication series

Name2019 International Symposium on VLSI Design, Automation and Test, VLSI-DAT 2019

Conference

Conference2019 International Symposium on VLSI Design, Automation and Test, VLSI-DAT 2019
CountryTaiwan
CityHsinchu
Period22/04/1925/04/19

Fingerprint Dive into the research topics of 'Energy efficient CNN inference accelerator using fast fourier transform'. Together they form a unique fingerprint.

Cite this