A digital circuit design of state-space recurrent neural networks

Che-Wei Lin; Jeen-Shing Wang

doi:10.1109/ICSMC.2008.4811556

A digital circuit design of state-space recurrent neural networks

Research output: Contribution to journal › Conference article › peer-review

Abstract

This paper presents a digital circuit design of a state-space recurrent neural network (RNN). The proposed digital circuit design separates the datapath of the state-space RNN into a linear subcircuit and a nonlinear subcircuit. The linear subcircuit is realized by a matrix-vector multiplier while the nonlinear subcircuit by a customized nonlinear function computing unit. The throughput rate of the proposed RNN circuit is 36060.5 times faster than that of the software simulation using MATLAB®. The proposed state-space RNN digital design methodology not only possesses the advantages including high computing speed, small area and portability, but also increases the possibility of using the digital RNN circuit in real-world dynamic problems.

Original language	English
Article number	4811556
Pages (from-to)	1838-1842
Number of pages	5
Journal	Conference Proceedings - IEEE International Conference on Systems, Man and Cybernetics
DOIs	https://doi.org/10.1109/ICSMC.2008.4811556
Publication status	Published - 2008 Dec 1
Event	2008 IEEE International Conference on Systems, Man and Cybernetics, SMC 2008 - Singapore, Singapore Duration: 2008 Oct 12 → 2008 Oct 15

All Science Journal Classification (ASJC) codes

Electrical and Electronic Engineering
Control and Systems Engineering
Human-Computer Interaction

Access to Document

10.1109/ICSMC.2008.4811556

Cite this

@article{3ad854fdfa2e4f3fadad507c5755f679,

title = "A digital circuit design of state-space recurrent neural networks",

abstract = "This paper presents a digital circuit design of a state-space recurrent neural network (RNN). The proposed digital circuit design separates the datapath of the state-space RNN into a linear subcircuit and a nonlinear subcircuit. The linear subcircuit is realized by a matrix-vector multiplier while the nonlinear subcircuit by a customized nonlinear function computing unit. The throughput rate of the proposed RNN circuit is 36060.5 times faster than that of the software simulation using MATLAB{\textregistered}. The proposed state-space RNN digital design methodology not only possesses the advantages including high computing speed, small area and portability, but also increases the possibility of using the digital RNN circuit in real-world dynamic problems.",

author = "Che-Wei Lin and Jeen-Shing Wang",

year = "2008",

month = dec,

day = "1",

doi = "10.1109/ICSMC.2008.4811556",

language = "English",

pages = "1838--1842",

journal = "Conference Proceedings - IEEE International Conference on Systems, Man and Cybernetics",

issn = "1062-922X",

note = "2008 IEEE International Conference on Systems, Man and Cybernetics, SMC 2008 ; Conference date: 12-10-2008 Through 15-10-2008",

}

TY - JOUR

T1 - A digital circuit design of state-space recurrent neural networks

AU - Lin, Che-Wei

AU - Wang, Jeen-Shing

PY - 2008/12/1

Y1 - 2008/12/1

N2 - This paper presents a digital circuit design of a state-space recurrent neural network (RNN). The proposed digital circuit design separates the datapath of the state-space RNN into a linear subcircuit and a nonlinear subcircuit. The linear subcircuit is realized by a matrix-vector multiplier while the nonlinear subcircuit by a customized nonlinear function computing unit. The throughput rate of the proposed RNN circuit is 36060.5 times faster than that of the software simulation using MATLAB®. The proposed state-space RNN digital design methodology not only possesses the advantages including high computing speed, small area and portability, but also increases the possibility of using the digital RNN circuit in real-world dynamic problems.

AB - This paper presents a digital circuit design of a state-space recurrent neural network (RNN). The proposed digital circuit design separates the datapath of the state-space RNN into a linear subcircuit and a nonlinear subcircuit. The linear subcircuit is realized by a matrix-vector multiplier while the nonlinear subcircuit by a customized nonlinear function computing unit. The throughput rate of the proposed RNN circuit is 36060.5 times faster than that of the software simulation using MATLAB®. The proposed state-space RNN digital design methodology not only possesses the advantages including high computing speed, small area and portability, but also increases the possibility of using the digital RNN circuit in real-world dynamic problems.

UR - http://www.scopus.com/inward/record.url?scp=69949122521&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=69949122521&partnerID=8YFLogxK

U2 - 10.1109/ICSMC.2008.4811556

DO - 10.1109/ICSMC.2008.4811556

M3 - Conference article

AN - SCOPUS:69949122521

SP - 1838

EP - 1842

JO - Conference Proceedings - IEEE International Conference on Systems, Man and Cybernetics

JF - Conference Proceedings - IEEE International Conference on Systems, Man and Cybernetics

SN - 1062-922X

M1 - 4811556

T2 - 2008 IEEE International Conference on Systems, Man and Cybernetics, SMC 2008

Y2 - 12 October 2008 through 15 October 2008

ER -

A digital circuit design of state-space recurrent neural networks

Abstract

All Science Journal Classification (ASJC) codes

Access to Document

Other files and links

Fingerprint

Cite this