Presentation 2018-06-14
Area Efficient Multiply-Accumulate Circuit Using Stochastic Computing for Neural Network Hardware
Kenta Nagura, Masayuki Hiromoto, Takashi Sato,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Neural network, which is an accurate and general-purpose machine learning method, is attracting greater attention in recent years. Due to the heavy computational load required in both learning and inference, the circuit area and power consumption become large when the neural network is implemented on a hardware. To improve calculation efficiency, we propose to apply stochastic computing (SC) in which the numerical numbers are represented by the number of 1's in a bit sequence. In this paper, we propose a new multiply-accumulate circuit (MAC) using SC, which is a heavily repeated calculation in neural network algorithms. Through experiments, we show the proposed MAC circuit greatly improves the accuracy of the calculation compared with an existing MAC circuit using SC of equal circuit area and power.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) neural network / multiply-accumulate circuit / image recognition / stochastic computing
Paper # CAS2018-15,VLD2018-18,SIP2018-35,MSS2018-15
Date of Issue 2018-06-07 (CAS, VLD, SIP, MSS)

Conference Information
Committee CAS / SIP / MSS / VLD
Conference Date 2018/6/14(2days)
Place (in Japanese) (See Japanese page)
Place (in English) Hokkaido Univ. (Frontier Research in Applied Sciences Build.)
Topics (in Japanese) (See Japanese page)
Topics (in English) System and Signal Processing, etc
Chair Hideaki Okazaki(Shonan Inst. of Tech.) / Shogo Muramatsu(Niigata Univ.) / Morikazu Nakamura(Univ. of Ryukyus) / Noriyuki Minegishi(Mitsubishi Electric)
Vice Chair Taizo Yamawaki(Hitachi) / Naoyuki Aikawa(TUS) / Kazunori Hayashi(Osaka City Univ) / Shigemasa Takai(Osaka Univ.) / Nozomu Togawa(Waseda Univ.)
Secretary Taizo Yamawaki(Shonan Inst. of Tech.) / Naoyuki Aikawa(Hitachi) / Kazunori Hayashi(Takushoku Univ.) / Shigemasa Takai(Hiroshima Univ.) / Nozomu Togawa(Toshiba)
Assistant Motoi Yamaguchi(Renesas Electronics) / / Hideki Kinjo(Okinawa Univ.)

Paper Information
Registration To Technical Committee on Circuits and Systems / Technical Committee on Signal Processing / Technical Committee on Mathematical Systems Science and its applications / Technical Committee on VLSI Design Technologies
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Area Efficient Multiply-Accumulate Circuit Using Stochastic Computing for Neural Network Hardware
Sub Title (in English)
Keyword(1) neural network
Keyword(2) multiply-accumulate circuit
Keyword(3) image recognition
Keyword(4) stochastic computing
1st Author's Name Kenta Nagura
1st Author's Affiliation Kyoto University(Kyoto Univ)
2nd Author's Name Masayuki Hiromoto
2nd Author's Affiliation Kyoto University(Kyoto Univ)
3rd Author's Name Takashi Sato
3rd Author's Affiliation Kyoto University(Kyoto Univ)
Date 2018-06-14
Paper # CAS2018-15,VLD2018-18,SIP2018-35,MSS2018-15
Volume (vol) vol.118
Number (no) CAS-82,VLD-83,SIP-84,MSS-85
Page pp.pp.81-86(CAS), pp.81-86(VLD), pp.81-86(SIP), pp.81-86(MSS),
#Pages 6
Date of Issue 2018-06-07 (CAS, VLD, SIP, MSS)