Presentation | 2021-03-03 Evaluation on Approximate Multiplier for CNN Calculation Yuechuan Zhang, Masahiro Fujita, Takashi Matsumoto, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | Improving the accuracy of a convolutional neural network (CNN) typically requires larger hardware with more energy consumption. On the other hand, the error tolerance of CNNs allows approximate computing to cut down the implementation costs. Given that multiplication is the most resource-intensive and power-hungry operation in CNNs, approximate multipliers (AMs) can be used to reduce hardware cost. There are various existing approximate multiplier generation methods. However, it remains unknown whether a specific AM is suitable for CNN calculation and its effect on inference accuracy. In this paper, we implement an 8-bit quantized Alexnet into hardware. The relationship between error caused by AMs and the accuracy of CNN inference is established. Mainly two AMs are considered: 1) Cartesian Genetic Programming (CGP)-based AMs 2) flexible Error-estimating based AMs. Besides, we propose two techniques, input zoom and error compensation to improve the performance of AMs in CNN calculation. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | CNN / Approximate multiplier / Error compensation |
Paper # | VLD2020-68,HWS2020-43 |
Date of Issue | 2021-02-24 (VLD, HWS) |
Conference Information | |
Committee | HWS / VLD |
---|---|
Conference Date | 2021/3/3(2days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | Online |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | Design Technology for System-on-Silicon, Hardware Security, etc. |
Chair | Makoto Ikeda(Univ. of Tokyo) / Daisuke Fukuda(Fujitsu Labs.) |
Vice Chair | Yasuhisa Shimazaki(Renesas Electronics) / Makoto Nagata(Kobe Univ.) / Kazutoshi Kobayashi(Kyoto Inst. of Tech.) |
Secretary | Yasuhisa Shimazaki(Kyushu Univ.) / Makoto Nagata(NTT) / Kazutoshi Kobayashi(Hitachi) |
Assistant | / Takuma Nishimoto(Hitachi) |
Paper Information | |
Registration To | Technical Committee on Hardware Security / Technical Committee on VLSI Design Technologies |
---|---|
Language | ENG |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Evaluation on Approximate Multiplier for CNN Calculation |
Sub Title (in English) | |
Keyword(1) | CNN |
Keyword(2) | Approximate multiplier |
Keyword(3) | Error compensation |
1st Author's Name | Yuechuan Zhang |
1st Author's Affiliation | Univerisity of Tokyo(UTokyo) |
2nd Author's Name | Masahiro Fujita |
2nd Author's Affiliation | Univerisity of Tokyo(UTokyo) |
3rd Author's Name | Takashi Matsumoto |
3rd Author's Affiliation | Univerisity of Tokyo(UTokyo) |
Date | 2021-03-03 |
Paper # | VLD2020-68,HWS2020-43 |
Volume (vol) | vol.120 |
Number (no) | VLD-400,HWS-401 |
Page | pp.pp.7-12(VLD), pp.7-12(HWS), |
#Pages | 6 |
Date of Issue | 2021-02-24 (VLD, HWS) |