A Reconfigurable Architecture for Implementing Locally Connected Neural Arrays

Research output: Chapter in Book/Report/Conference proceedingConference contribution

69 Downloads (Pure)

Abstract

Moore’s law is rapidly approaching a long-predicted decline, and with it the performance gains of conventional processors are becoming ever more marginal. Cognitive computing systems based on neural networks have the potential to provide a solution to the decline of Moore’s law. Identifying common traits in neural systems can lead to the design of more efficient, robust and adaptable processors. Despite the potentials, large-scale neural systems remain difficult to implement due to constraints on scalability. Here we introduce a new hardware architecture for implementing locally connected neural networks that can model biological systems with a high level of scalability. We validate our architecture using a full model of the locomotion system of the Caenorhabditis elegans. Further, we show that our proposed architecture archives a nine-fold increase in clock speed over existing hardware models. Importantly the clock speed for our architecture is found to be independent of system size, providing an unparalleled level of scalability. Our approach can be applied to the modelling of large neural networks, with greater performance, easier configuration and a high level of scalability.
Original languageEnglish
Title of host publicationProceedings of the 2018 Science and Information (SAI) Computing Conference
EditorsK. Arai, S. Kapoor, R. Bhatia
PublisherCurran Associates
Pages76-92
Number of pages17
ISBN (Print)9783030011765
DOIs
Publication statusAccepted/In press - 26 Apr 2018
EventComputing Conference 2018 - London, UK United Kingdom
Duration: 10 Jul 201812 Jul 2018
http://saiconference.com/Computing

Publication series

NameAdvances in Intelligent Systems and Computing
Volume857

Conference

ConferenceComputing Conference 2018
CountryUK United Kingdom
CityLondon
Period10/07/1812/07/18
Internet address

Keywords

  • Caenorhabditis elegans
  • Neural network
  • Neuromorphic
  • PNAA
  • Reconfigurable hardware

ASJC Scopus subject areas

  • Control and Systems Engineering
  • Computer Science(all)

Cite this

Graham-Harper-Cater, J., Clarke, C., Metcalfe, B., & Wilson, P. (Accepted/In press). A Reconfigurable Architecture for Implementing Locally Connected Neural Arrays. In K. Arai, S. Kapoor, & R. Bhatia (Eds.), Proceedings of the 2018 Science and Information (SAI) Computing Conference (pp. 76-92). (Advances in Intelligent Systems and Computing; Vol. 857). Curran Associates. https://doi.org/10.1007/978-3-030-01177-2_6

A Reconfigurable Architecture for Implementing Locally Connected Neural Arrays. / Graham-Harper-Cater, Jonathan; Clarke, Christopher; Metcalfe, Benjamin; Wilson, Peter.

Proceedings of the 2018 Science and Information (SAI) Computing Conference. ed. / K. Arai; S. Kapoor; R. Bhatia. Curran Associates, 2018. p. 76-92 (Advances in Intelligent Systems and Computing; Vol. 857).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Graham-Harper-Cater, J, Clarke, C, Metcalfe, B & Wilson, P 2018, A Reconfigurable Architecture for Implementing Locally Connected Neural Arrays. in K Arai, S Kapoor & R Bhatia (eds), Proceedings of the 2018 Science and Information (SAI) Computing Conference. Advances in Intelligent Systems and Computing, vol. 857, Curran Associates, pp. 76-92, Computing Conference 2018, London, UK United Kingdom, 10/07/18. https://doi.org/10.1007/978-3-030-01177-2_6
Graham-Harper-Cater J, Clarke C, Metcalfe B, Wilson P. A Reconfigurable Architecture for Implementing Locally Connected Neural Arrays. In Arai K, Kapoor S, Bhatia R, editors, Proceedings of the 2018 Science and Information (SAI) Computing Conference. Curran Associates. 2018. p. 76-92. (Advances in Intelligent Systems and Computing). https://doi.org/10.1007/978-3-030-01177-2_6
Graham-Harper-Cater, Jonathan ; Clarke, Christopher ; Metcalfe, Benjamin ; Wilson, Peter. / A Reconfigurable Architecture for Implementing Locally Connected Neural Arrays. Proceedings of the 2018 Science and Information (SAI) Computing Conference. editor / K. Arai ; S. Kapoor ; R. Bhatia. Curran Associates, 2018. pp. 76-92 (Advances in Intelligent Systems and Computing).
@inproceedings{e1f50f51f5384eaf8f20b8ac2c74e030,
title = "A Reconfigurable Architecture for Implementing Locally Connected Neural Arrays",
abstract = "Moore’s law is rapidly approaching a long-predicted decline, and with it the performance gains of conventional processors are becoming ever more marginal. Cognitive computing systems based on neural networks have the potential to provide a solution to the decline of Moore’s law. Identifying common traits in neural systems can lead to the design of more efficient, robust and adaptable processors. Despite the potentials, large-scale neural systems remain difficult to implement due to constraints on scalability. Here we introduce a new hardware architecture for implementing locally connected neural networks that can model biological systems with a high level of scalability. We validate our architecture using a full model of the locomotion system of the Caenorhabditis elegans. Further, we show that our proposed architecture archives a nine-fold increase in clock speed over existing hardware models. Importantly the clock speed for our architecture is found to be independent of system size, providing an unparalleled level of scalability. Our approach can be applied to the modelling of large neural networks, with greater performance, easier configuration and a high level of scalability.",
keywords = "Caenorhabditis elegans, Neural network, Neuromorphic, PNAA, Reconfigurable hardware",
author = "Jonathan Graham-Harper-Cater and Christopher Clarke and Benjamin Metcalfe and Peter Wilson",
year = "2018",
month = "4",
day = "26",
doi = "10.1007/978-3-030-01177-2_6",
language = "English",
isbn = "9783030011765",
series = "Advances in Intelligent Systems and Computing",
publisher = "Curran Associates",
pages = "76--92",
editor = "K. Arai and S. Kapoor and R. Bhatia",
booktitle = "Proceedings of the 2018 Science and Information (SAI) Computing Conference",

}

TY - GEN

T1 - A Reconfigurable Architecture for Implementing Locally Connected Neural Arrays

AU - Graham-Harper-Cater, Jonathan

AU - Clarke, Christopher

AU - Metcalfe, Benjamin

AU - Wilson, Peter

PY - 2018/4/26

Y1 - 2018/4/26

N2 - Moore’s law is rapidly approaching a long-predicted decline, and with it the performance gains of conventional processors are becoming ever more marginal. Cognitive computing systems based on neural networks have the potential to provide a solution to the decline of Moore’s law. Identifying common traits in neural systems can lead to the design of more efficient, robust and adaptable processors. Despite the potentials, large-scale neural systems remain difficult to implement due to constraints on scalability. Here we introduce a new hardware architecture for implementing locally connected neural networks that can model biological systems with a high level of scalability. We validate our architecture using a full model of the locomotion system of the Caenorhabditis elegans. Further, we show that our proposed architecture archives a nine-fold increase in clock speed over existing hardware models. Importantly the clock speed for our architecture is found to be independent of system size, providing an unparalleled level of scalability. Our approach can be applied to the modelling of large neural networks, with greater performance, easier configuration and a high level of scalability.

AB - Moore’s law is rapidly approaching a long-predicted decline, and with it the performance gains of conventional processors are becoming ever more marginal. Cognitive computing systems based on neural networks have the potential to provide a solution to the decline of Moore’s law. Identifying common traits in neural systems can lead to the design of more efficient, robust and adaptable processors. Despite the potentials, large-scale neural systems remain difficult to implement due to constraints on scalability. Here we introduce a new hardware architecture for implementing locally connected neural networks that can model biological systems with a high level of scalability. We validate our architecture using a full model of the locomotion system of the Caenorhabditis elegans. Further, we show that our proposed architecture archives a nine-fold increase in clock speed over existing hardware models. Importantly the clock speed for our architecture is found to be independent of system size, providing an unparalleled level of scalability. Our approach can be applied to the modelling of large neural networks, with greater performance, easier configuration and a high level of scalability.

KW - Caenorhabditis elegans

KW - Neural network

KW - Neuromorphic

KW - PNAA

KW - Reconfigurable hardware

UR - http://www.scopus.com/inward/record.url?scp=85057118243&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-01177-2_6

DO - 10.1007/978-3-030-01177-2_6

M3 - Conference contribution

SN - 9783030011765

T3 - Advances in Intelligent Systems and Computing

SP - 76

EP - 92

BT - Proceedings of the 2018 Science and Information (SAI) Computing Conference

A2 - Arai, K.

A2 - Kapoor, S.

A2 - Bhatia, R.

PB - Curran Associates

ER -