Optimal mapping of neural networks onto FPGAs1 - A new constructive algorithm

Valeriu Beiu; John G. Taylor

doi:10.1007/3-540-59497-3_256

Optimal mapping of neural networks onto FPGAs¹ - A new constructive algorithm

Valeriu Beiu, John G. Taylor

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

3 Citations (Scopus)

Abstract

The paper shows how a feedforward neural network defined by a set of m binary examples of n bits each, can be determined by a novel constructive algorithm (which determines the number of layers, the number of neurons in each layer and the synaptic weights of a particular neural network). For doing that, the optimisation criteria of the new algorithm can be chosen from the following: (i) the area of the circuit A; (ii) the AT² complexity measure of VLSI; (iii) the delay T; or (iv) the maximum fan-in for a gate A. As a result the neural network which is build can be optimised for mapping onto a FPGA. By considering the maximum fan-in of one neuron as a parameter, we proceed to show its influence on the area, and suggest how to obtain a full class of solutions. We also compare our results with other constructive algorithms and benchmark it on the classical “two spirals problem.” Conclusions and some open problems are closing the paper.

Original language	English
Title of host publication	From Natural to Artificial Neural Computation - International Workshop on Artificial Neural Networks, Proceedings
Editors	Jose Mira, Francisco Sandoval
Publisher	Springer Verlag
Pages	822-829
Number of pages	8
ISBN (Print)	3540594973, 9783540594970
DOIs	https://doi.org/10.1007/3-540-59497-3_256
Publication status	Published - 1995
Externally published	Yes
Event	3rd International Workshop on Artificial Neural Networks, IWANN 1995 - Malaga-Torremolinos, Spain Duration: Jun 7 1995 → Jun 9 1995

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	930
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Other

Other	3rd International Workshop on Artificial Neural Networks, IWANN 1995
Country/Territory	Spain
City	Malaga-Torremolinos
Period	6/7/95 → 6/9/95

ASJC Scopus subject areas

Theoretical Computer Science
General Computer Science

Access to Document

10.1007/3-540-59497-3_256

Cite this

Beiu, V., & Taylor, J. G. (1995). Optimal mapping of neural networks onto FPGAs¹ - A new constructive algorithm. In J. Mira, & F. Sandoval (Eds.), From Natural to Artificial Neural Computation - International Workshop on Artificial Neural Networks, Proceedings (pp. 822-829). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 930). Springer Verlag. https://doi.org/10.1007/3-540-59497-3_256

Optimal mapping of neural networks onto FPGAs¹ - A new constructive algorithm. / Beiu, Valeriu; Taylor, John G.
From Natural to Artificial Neural Computation - International Workshop on Artificial Neural Networks, Proceedings. ed. / Jose Mira; Francisco Sandoval. Springer Verlag, 1995. p. 822-829 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 930).

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

Beiu, V & Taylor, JG 1995, Optimal mapping of neural networks onto FPGAs¹ - A new constructive algorithm. in J Mira & F Sandoval (eds), From Natural to Artificial Neural Computation - International Workshop on Artificial Neural Networks, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 930, Springer Verlag, pp. 822-829, 3rd International Workshop on Artificial Neural Networks, IWANN 1995, Malaga-Torremolinos, Spain, 6/7/95. https://doi.org/10.1007/3-540-59497-3_256

Beiu V, Taylor JG. Optimal mapping of neural networks onto FPGAs¹ - A new constructive algorithm. In Mira J, Sandoval F, editors, From Natural to Artificial Neural Computation - International Workshop on Artificial Neural Networks, Proceedings. Springer Verlag. 1995. p. 822-829. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/3-540-59497-3_256

Beiu, Valeriu ; Taylor, John G. / Optimal mapping of neural networks onto FPGAs¹ - A new constructive algorithm. From Natural to Artificial Neural Computation - International Workshop on Artificial Neural Networks, Proceedings. editor / Jose Mira ; Francisco Sandoval. Springer Verlag, 1995. pp. 822-829 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{19c32aa5471b4243a906ce865e8f27ad,

title = "Optimal mapping of neural networks onto FPGAs1 - A new constructive algorithm",

abstract = "The paper shows how a feedforward neural network defined by a set of m binary examples of n bits each, can be determined by a novel constructive algorithm (which determines the number of layers, the number of neurons in each layer and the synaptic weights of a particular neural network). For doing that, the optimisation criteria of the new algorithm can be chosen from the following: (i) the area of the circuit A; (ii) the AT2 complexity measure of VLSI; (iii) the delay T; or (iv) the maximum fan-in for a gate A. As a result the neural network which is build can be optimised for mapping onto a FPGA. By considering the maximum fan-in of one neuron as a parameter, we proceed to show its influence on the area, and suggest how to obtain a full class of solutions. We also compare our results with other constructive algorithms and benchmark it on the classical “two spirals problem.” Conclusions and some open problems are closing the paper.",

author = "Valeriu Beiu and Taylor, {John G.}",

note = "Publisher Copyright: {\textcopyright} 1995, Springer Verlag. All rights reserved.; 3rd International Workshop on Artificial Neural Networks, IWANN 1995 ; Conference date: 07-06-1995 Through 09-06-1995",

year = "1995",

doi = "10.1007/3-540-59497-3_256",

language = "English",

isbn = "3540594973",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag",

pages = "822--829",

editor = "Jose Mira and Francisco Sandoval",

booktitle = "From Natural to Artificial Neural Computation - International Workshop on Artificial Neural Networks, Proceedings",

}

TY - GEN

T1 - Optimal mapping of neural networks onto FPGAs1 - A new constructive algorithm

AU - Beiu, Valeriu

AU - Taylor, John G.

PY - 1995

Y1 - 1995

N2 - The paper shows how a feedforward neural network defined by a set of m binary examples of n bits each, can be determined by a novel constructive algorithm (which determines the number of layers, the number of neurons in each layer and the synaptic weights of a particular neural network). For doing that, the optimisation criteria of the new algorithm can be chosen from the following: (i) the area of the circuit A; (ii) the AT2 complexity measure of VLSI; (iii) the delay T; or (iv) the maximum fan-in for a gate A. As a result the neural network which is build can be optimised for mapping onto a FPGA. By considering the maximum fan-in of one neuron as a parameter, we proceed to show its influence on the area, and suggest how to obtain a full class of solutions. We also compare our results with other constructive algorithms and benchmark it on the classical “two spirals problem.” Conclusions and some open problems are closing the paper.

AB - The paper shows how a feedforward neural network defined by a set of m binary examples of n bits each, can be determined by a novel constructive algorithm (which determines the number of layers, the number of neurons in each layer and the synaptic weights of a particular neural network). For doing that, the optimisation criteria of the new algorithm can be chosen from the following: (i) the area of the circuit A; (ii) the AT2 complexity measure of VLSI; (iii) the delay T; or (iv) the maximum fan-in for a gate A. As a result the neural network which is build can be optimised for mapping onto a FPGA. By considering the maximum fan-in of one neuron as a parameter, we proceed to show its influence on the area, and suggest how to obtain a full class of solutions. We also compare our results with other constructive algorithms and benchmark it on the classical “two spirals problem.” Conclusions and some open problems are closing the paper.

UR - http://www.scopus.com/inward/record.url?scp=0344828787&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0344828787&partnerID=8YFLogxK

U2 - 10.1007/3-540-59497-3_256

DO - 10.1007/3-540-59497-3_256

M3 - Conference contribution

AN - SCOPUS:0344828787

SN - 3540594973

SN - 9783540594970

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 822

EP - 829

BT - From Natural to Artificial Neural Computation - International Workshop on Artificial Neural Networks, Proceedings

A2 - Mira, Jose

A2 - Sandoval, Francisco

PB - Springer Verlag

T2 - 3rd International Workshop on Artificial Neural Networks, IWANN 1995

Y2 - 7 June 1995 through 9 June 1995

ER -