A Prototype 3D Optically Interconnected Neural Network

GöKçe Yayla, Sadik Esener, Ashok V. Krishnamoorthy, Gary C. Marsden

Research output: Contribution to journalArticle

14 Citations (Scopus)

Abstract

We report the implementation of a prototype three-dimensional (3D) optoelectronic neural network that combines free-space optical interconnects with silicon-VLSI-based optoelectronic circuits. The prototype system consists of a 16-node input, 4-neuron hidden, and a single-neuron output layer, where the denser input-to-hidden-layer connections are optical. The input layer uses PLZT light modulators to generate optical outputs which are distributed over an optoelectronic neural network chip through space-invariant holographic optical interconnects. Optical interconnections provide negligible fan-out delay and allow compact, purely on-chip electronic H-tree type fan-in structure. The small prototype system achieves a measured 8-bit electronic fan-in precision and a calculated maximum speed of 640 million interconnections per second. The system was tested using synaptic weights learned off system and was shown to distinguish any vertical line from any horizontal one in an image of 4/spl times/4 pixels. New, more efficient light detector and small-area analog synapse circuits and denser optoelectronic neuron layouts are proposed to scale up the system. A high-speed, feed-forward optoelectronic synapse implementation density of up to 10/sup 4//cm/sup 2/ seems feasible using new synapse design. A scaling analysis of the system shows that the optically interconnected neural network implementation can provide higher fan-in speed and lower power consumption characteristics than a purely electronic, crossbar-based neural network implementation.

Original languageEnglish (US)
Pages (from-to)1749-1762
Number of pages14
JournalProceedings of the IEEE
Volume82
Issue number11
DOIs
StatePublished - 1994
Externally publishedYes

Fingerprint

Optoelectronic devices
Optical interconnects
Neural networks
Neurons
Networks (circuits)
Light modulators
Photodetectors
Fans
Electric power utilization
Pixels
Silicon

ASJC Scopus subject areas

  • Electrical and Electronic Engineering

Cite this

Yayla, G., Esener, S., Krishnamoorthy, A. V., & Marsden, G. C. (1994). A Prototype 3D Optically Interconnected Neural Network. Proceedings of the IEEE, 82(11), 1749-1762. https://doi.org/10.1109/5.333752

A Prototype 3D Optically Interconnected Neural Network. / Yayla, GöKçe; Esener, Sadik; Krishnamoorthy, Ashok V.; Marsden, Gary C.

In: Proceedings of the IEEE, Vol. 82, No. 11, 1994, p. 1749-1762.

Research output: Contribution to journalArticle

Yayla, G, Esener, S, Krishnamoorthy, AV & Marsden, GC 1994, 'A Prototype 3D Optically Interconnected Neural Network', Proceedings of the IEEE, vol. 82, no. 11, pp. 1749-1762. https://doi.org/10.1109/5.333752
Yayla, GöKçe ; Esener, Sadik ; Krishnamoorthy, Ashok V. ; Marsden, Gary C. / A Prototype 3D Optically Interconnected Neural Network. In: Proceedings of the IEEE. 1994 ; Vol. 82, No. 11. pp. 1749-1762.
@article{2a6038f7f3124e3592215001a65e824d,
title = "A Prototype 3D Optically Interconnected Neural Network",
abstract = "We report the implementation of a prototype three-dimensional (3D) optoelectronic neural network that combines free-space optical interconnects with silicon-VLSI-based optoelectronic circuits. The prototype system consists of a 16-node input, 4-neuron hidden, and a single-neuron output layer, where the denser input-to-hidden-layer connections are optical. The input layer uses PLZT light modulators to generate optical outputs which are distributed over an optoelectronic neural network chip through space-invariant holographic optical interconnects. Optical interconnections provide negligible fan-out delay and allow compact, purely on-chip electronic H-tree type fan-in structure. The small prototype system achieves a measured 8-bit electronic fan-in precision and a calculated maximum speed of 640 million interconnections per second. The system was tested using synaptic weights learned off system and was shown to distinguish any vertical line from any horizontal one in an image of 4/spl times/4 pixels. New, more efficient light detector and small-area analog synapse circuits and denser optoelectronic neuron layouts are proposed to scale up the system. A high-speed, feed-forward optoelectronic synapse implementation density of up to 10/sup 4//cm/sup 2/ seems feasible using new synapse design. A scaling analysis of the system shows that the optically interconnected neural network implementation can provide higher fan-in speed and lower power consumption characteristics than a purely electronic, crossbar-based neural network implementation.",
author = "G{\"o}K{\cc}e Yayla and Sadik Esener and Krishnamoorthy, {Ashok V.} and Marsden, {Gary C.}",
year = "1994",
doi = "10.1109/5.333752",
language = "English (US)",
volume = "82",
pages = "1749--1762",
journal = "Proceedings of the IEEE",
issn = "0018-9219",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "11",

}

TY - JOUR

T1 - A Prototype 3D Optically Interconnected Neural Network

AU - Yayla, GöKçe

AU - Esener, Sadik

AU - Krishnamoorthy, Ashok V.

AU - Marsden, Gary C.

PY - 1994

Y1 - 1994

N2 - We report the implementation of a prototype three-dimensional (3D) optoelectronic neural network that combines free-space optical interconnects with silicon-VLSI-based optoelectronic circuits. The prototype system consists of a 16-node input, 4-neuron hidden, and a single-neuron output layer, where the denser input-to-hidden-layer connections are optical. The input layer uses PLZT light modulators to generate optical outputs which are distributed over an optoelectronic neural network chip through space-invariant holographic optical interconnects. Optical interconnections provide negligible fan-out delay and allow compact, purely on-chip electronic H-tree type fan-in structure. The small prototype system achieves a measured 8-bit electronic fan-in precision and a calculated maximum speed of 640 million interconnections per second. The system was tested using synaptic weights learned off system and was shown to distinguish any vertical line from any horizontal one in an image of 4/spl times/4 pixels. New, more efficient light detector and small-area analog synapse circuits and denser optoelectronic neuron layouts are proposed to scale up the system. A high-speed, feed-forward optoelectronic synapse implementation density of up to 10/sup 4//cm/sup 2/ seems feasible using new synapse design. A scaling analysis of the system shows that the optically interconnected neural network implementation can provide higher fan-in speed and lower power consumption characteristics than a purely electronic, crossbar-based neural network implementation.

AB - We report the implementation of a prototype three-dimensional (3D) optoelectronic neural network that combines free-space optical interconnects with silicon-VLSI-based optoelectronic circuits. The prototype system consists of a 16-node input, 4-neuron hidden, and a single-neuron output layer, where the denser input-to-hidden-layer connections are optical. The input layer uses PLZT light modulators to generate optical outputs which are distributed over an optoelectronic neural network chip through space-invariant holographic optical interconnects. Optical interconnections provide negligible fan-out delay and allow compact, purely on-chip electronic H-tree type fan-in structure. The small prototype system achieves a measured 8-bit electronic fan-in precision and a calculated maximum speed of 640 million interconnections per second. The system was tested using synaptic weights learned off system and was shown to distinguish any vertical line from any horizontal one in an image of 4/spl times/4 pixels. New, more efficient light detector and small-area analog synapse circuits and denser optoelectronic neuron layouts are proposed to scale up the system. A high-speed, feed-forward optoelectronic synapse implementation density of up to 10/sup 4//cm/sup 2/ seems feasible using new synapse design. A scaling analysis of the system shows that the optically interconnected neural network implementation can provide higher fan-in speed and lower power consumption characteristics than a purely electronic, crossbar-based neural network implementation.

UR - http://www.scopus.com/inward/record.url?scp=0028546998&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0028546998&partnerID=8YFLogxK

U2 - 10.1109/5.333752

DO - 10.1109/5.333752

M3 - Article

AN - SCOPUS:0028546998

VL - 82

SP - 1749

EP - 1762

JO - Proceedings of the IEEE

JF - Proceedings of the IEEE

SN - 0018-9219

IS - 11

ER -