Title:
NEURAL NETWORK ARITHMETIC CIRCUIT USING NON-VOLATILE SEMICONDUCTOR MEMORY ELEMENT
Document Type and Number:
WIPO Patent Application WO/2019/049741
Kind Code:
A1
Abstract:
A neural network arithmetic circuit for outputting output data (y) in accordance with the results of a multiply-add operation performed on input data (x0-xn) and connection weight coefficients (w0-wn), wherein the neural network arithmetic circuit is provided with arithmetic units (PU0-PUn) each comprising a non-volatile semiconductor memory element (RP) and cell transistor (T0) connected in series between data lines (BL0, SL0), a non-volatile semiconductor memory element (RN) and cell transistor (T1) connected in series between data lines (BL1, SL1), and a word line (WL0-WLn) connected to the gates of the cell transistors (T0, T1), connection weight coefficients (w0-wn) are stored in the RPs and Rns, a word line selection circuit (30) sets WL0-WLn to a selected state or non-selected state in accordance with xo-xn, and a determination circuit (50) determines the values of the currents flowing through BL0 and BL1 and accordingly outputs the output data (y).
Inventors:
KOUNO KAZUYUKI
ONO TAKASHI
NAKAYAMA MASAYOSHI
MOCHIDA REIJI
HAYATA YURIKO
ONO TAKASHI
NAKAYAMA MASAYOSHI
MOCHIDA REIJI
HAYATA YURIKO
Application Number:
PCT/JP2018/031899
Publication Date:
March 14, 2019
Filing Date:
August 29, 2018
Export Citation:
Assignee:
PANASONIC CORP (JP)
International Classes:
G11C11/54; G06G7/60; G06N3/063
Foreign References:
| US20140122402A1 | 2014-05-01 | |||
| JP2010146514A | 2010-07-01 | |||
| JP2003283003A | 2003-10-03 | |||
| JP2001188767A | 2001-07-10 | |||
| JPH06259585A | 1994-09-16 | |||
| JPH02161556A | 1990-06-21 |
Other References:
M. PREZIOSO ET AL.: "Training and operation of an integrated neuromorphic network based on metal-oxide memristors", NATURE, 2015, pages 61 - 64, XP055568356, DOI: 10.1038/nature14441
Attorney, Agent or Firm:
NII, Hiromori et al. (JP)
Download PDF: