Presentation 1998/4/23
A Self-Learning Digital Neural Network LSI using Sparse-Memory-Access Architecture
Osamu Saito, Kimihisa Aihara, Osamu Fujita, Kuniharu Uchimura,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) This digital neural network chip executes 10 GCPS for forward clalculations and 1 GCUPS for learning calculations. The chip can handle a 1 million synapse network together with external SRAM chips. Using sparse memory-access (SMA) architecture, unnecessary operations due to the nonlinera nature of neuron can be automatically eliminated without an effect on accuracy. A chip is fabricated using a 0.25 μm CMOS embedded array process and contains 64 processing units (UPs), and an eight-port external RAM access control unit to take full advantage of SMA architecute.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Neural Network / Neurochip / Digital Signal Processing / VLSI
Paper #
Date of Issue

Conference Information
Committee ICD
Conference Date 1998/4/23(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Integrated Circuits and Devices (ICD)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) A Self-Learning Digital Neural Network LSI using Sparse-Memory-Access Architecture
Sub Title (in English)
Keyword(1) Neural Network
Keyword(2) Neurochip
Keyword(3) Digital Signal Processing
Keyword(4) VLSI
1st Author's Name Osamu Saito
1st Author's Affiliation NTT Integrated Information & Energy Systems Laboratories()
2nd Author's Name Kimihisa Aihara
2nd Author's Affiliation NTT Network Service Systems Laboratories
3rd Author's Name Osamu Fujita
3rd Author's Affiliation NTT System Elecrtronics Laboratories
4th Author's Name Kuniharu Uchimura
4th Author's Affiliation NTT Integrated Information & Energy Systems Laboratories
Date 1998/4/23
Paper #
Volume (vol) vol.98
Number (no) 22
Page pp.pp.-
#Pages 8
Date of Issue