Defect tolerant implementations of feed-forward and recurrent neural networks

Paul Franzon; David Van den Bout; John Paulos; Thomas Miller; Wesley Snyder; Troy Nagle; Wentai Liu

Defect tolerant implementations of feed-forward and recurrent neural networks

Paul Franzon, David Van den Bout, John Paulos, Thomas Miller, Wesley Snyder, Troy Nagle, Wentai Liu

Department of Biomedical Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

The issues involved in the defect tolerant, large-scale implementation of two basic architectural classes of artificial neural networks--feed-forward and recurrent--are considered. The structures can be reconfigured to enable differently sized and connected neural structures to be implemented on the same piece of hardware. With the addition of suitable test measures, the techniques that give scalable and flexible neural networks also give defect tolerance. Thus the advantages of wafer-scale integration (WSI) can be readily applied to these structures. By applying WSI techniques, scalability is enhanced, as the requirement for the network to be partitioned between chips can be greatly relaxed. Thus, richly connected networks can be constructed.

Original language	English
Title of host publication	1990 Proc Int Conf Wafer Scale Integr
Editors	Joe Brewer, Michael J. Little
Publisher	Publ by IEEE
Pages	160-166
Number of pages	7
ISBN (Print)	0818690135
Publication status	Published - 1990
Event	1990 Proceedings - International Conference on Wafer Scale Integration - San Francisco, CA, USA Duration: 1990 Jan 23 → 1990 Jan 25

Publication series

Name	1990 Proc Int Conf Wafer Scale Integr

Conference

Conference	1990 Proceedings - International Conference on Wafer Scale Integration
City	San Francisco, CA, USA
Period	90-01-23 → 90-01-25

All Science Journal Classification (ASJC) codes

General Engineering

Cite this

@inproceedings{f249c138e6194b9b9ebdae46e6e997b5,

title = "Defect tolerant implementations of feed-forward and recurrent neural networks",

abstract = "The issues involved in the defect tolerant, large-scale implementation of two basic architectural classes of artificial neural networks--feed-forward and recurrent--are considered. The structures can be reconfigured to enable differently sized and connected neural structures to be implemented on the same piece of hardware. With the addition of suitable test measures, the techniques that give scalable and flexible neural networks also give defect tolerance. Thus the advantages of wafer-scale integration (WSI) can be readily applied to these structures. By applying WSI techniques, scalability is enhanced, as the requirement for the network to be partitioned between chips can be greatly relaxed. Thus, richly connected networks can be constructed.",

author = "Paul Franzon and {Van den Bout}, David and John Paulos and Thomas Miller and Wesley Snyder and Troy Nagle and Wentai Liu",

year = "1990",

language = "English",

isbn = "0818690135",

series = "1990 Proc Int Conf Wafer Scale Integr",

publisher = "Publ by IEEE",

pages = "160--166",

editor = "Joe Brewer and Little, {Michael J.}",

booktitle = "1990 Proc Int Conf Wafer Scale Integr",

note = "1990 Proceedings - International Conference on Wafer Scale Integration ; Conference date: 23-01-1990 Through 25-01-1990",

}

Franzon, P, Van den Bout, D, Paulos, J, Miller, T, Snyder, W, Nagle, T & Liu, W 1990, Defect tolerant implementations of feed-forward and recurrent neural networks. in J Brewer & MJ Little (eds), 1990 Proc Int Conf Wafer Scale Integr. 1990 Proc Int Conf Wafer Scale Integr, Publ by IEEE, pp. 160-166, 1990 Proceedings - International Conference on Wafer Scale Integration, San Francisco, CA, USA, 90-01-23.

Defect tolerant implementations of feed-forward and recurrent neural networks. / Franzon, Paul; Van den Bout, David; Paulos, John et al.
1990 Proc Int Conf Wafer Scale Integr. ed. / Joe Brewer; Michael J. Little. Publ by IEEE, 1990. p. 160-166 (1990 Proc Int Conf Wafer Scale Integr).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - Defect tolerant implementations of feed-forward and recurrent neural networks

AU - Franzon, Paul

AU - Van den Bout, David

AU - Paulos, John

AU - Miller, Thomas

AU - Snyder, Wesley

AU - Nagle, Troy

AU - Liu, Wentai

PY - 1990

Y1 - 1990

N2 - The issues involved in the defect tolerant, large-scale implementation of two basic architectural classes of artificial neural networks--feed-forward and recurrent--are considered. The structures can be reconfigured to enable differently sized and connected neural structures to be implemented on the same piece of hardware. With the addition of suitable test measures, the techniques that give scalable and flexible neural networks also give defect tolerance. Thus the advantages of wafer-scale integration (WSI) can be readily applied to these structures. By applying WSI techniques, scalability is enhanced, as the requirement for the network to be partitioned between chips can be greatly relaxed. Thus, richly connected networks can be constructed.

AB - The issues involved in the defect tolerant, large-scale implementation of two basic architectural classes of artificial neural networks--feed-forward and recurrent--are considered. The structures can be reconfigured to enable differently sized and connected neural structures to be implemented on the same piece of hardware. With the addition of suitable test measures, the techniques that give scalable and flexible neural networks also give defect tolerance. Thus the advantages of wafer-scale integration (WSI) can be readily applied to these structures. By applying WSI techniques, scalability is enhanced, as the requirement for the network to be partitioned between chips can be greatly relaxed. Thus, richly connected networks can be constructed.

UR - http://www.scopus.com/inward/record.url?scp=0025248950&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0025248950&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:0025248950

SN - 0818690135

T3 - 1990 Proc Int Conf Wafer Scale Integr

SP - 160

EP - 166

BT - 1990 Proc Int Conf Wafer Scale Integr

A2 - Brewer, Joe

A2 - Little, Michael J.

PB - Publ by IEEE

T2 - 1990 Proceedings - International Conference on Wafer Scale Integration

Y2 - 23 January 1990 through 25 January 1990

ER -

Defect tolerant implementations of feed-forward and recurrent neural networks

Abstract

Publication series

Conference

All Science Journal Classification (ASJC) codes

Other files and links

Fingerprint

Cite this