Presentation 1993/9/17
A Study on High-precision Analog Neural Network VLSI Computers
Takashi Morie, Yoshihito Amemiya,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) This paper demonstrates usefulness of analog neuro-LSIs to construct general neural networks with analog dynamics.From a perspective of calculation resolution,analog LSI implementation is more suitable for high speed neuraj networks than digital LSI implementation.It is also demonstrated that noise in analog circuits less affects the learning performance than weight quantization in digital memories.It is described that serious effects of offset errors in analog circuits on the backpropagation(BP)learning performance can be lowered by using a new learning algorithm called contrastive BP learning,and a circuit architecture implementing the learning procedure is explained.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) neural network / analog LSI / backpropagation learning / contrastive learning
Paper # ICD93-97,DSP93-58
Date of Issue

Conference Information
Committee ICD
Conference Date 1993/9/17(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 Study on High-precision Analog Neural Network VLSI Computers
Sub Title (in English)
Keyword(1) neural network
Keyword(2) analog LSI
Keyword(3) backpropagation learning
Keyword(4) contrastive learning
1st Author's Name Takashi Morie
1st Author's Affiliation NTT LSI Laboratories()
2nd Author's Name Yoshihito Amemiya
2nd Author's Affiliation Faculty of Engineering,Hokkaido University
Date 1993/9/17
Paper # ICD93-97,DSP93-58
Volume (vol) vol.93
Number (no) 231
Page pp.pp.-
#Pages 8
Date of Issue