Event-driven implementation of deep spiking convolutional neural networks for supervised classification using the SpiNNaker neuromorphic platform

Neural networks have enabled great advances in recent times due mainly to improved parallel computing capabilities in accordance to Moore’s Law, which allowed reducing the time needed for the parameter learning of complex, multi-layered neural architectures. However, with silicon technology reaching...

ver descrição completa

Detalhes bibliográficos
Autores: Patiño Saucedo, Alberto, Rostro González, Horacio, Serrano Gotarredona, María Teresa, Linares Barranco, Bernabé
Formato: artículo
Estado:Versión enviada para evaluación y publicación
Fecha de publicación:2020
País:España
Recursos:Universidad de Sevilla (US)
Repositorio:idUS. Depósito de Investigación de la Universidad de Sevilla
OAI Identifier:oai:idus.us.es:11441/102108
Acesso em linha:https://hdl.handle.net/11441/102108
https://doi.org/10.1016/j.neunet.2019.09.008
Access Level:acceso abierto
Palavra-chave:Neuromorphic hardware
Artificial neural networks
Spiking neural networks
MNIST
SpiNNaker
Event processing
id ES_60dff3695329bb93066a2b0105cb542f
oai_identifier_str oai:idus.us.es:11441/102108
network_acronym_str ES
network_name_str España
repository_id_str
spelling Event-driven implementation of deep spiking convolutional neural networks for supervised classification using the SpiNNaker neuromorphic platformPatiño Saucedo, AlbertoRostro González, HoracioSerrano Gotarredona, María TeresaLinares Barranco, BernabéNeuromorphic hardwareArtificial neural networksSpiking neural networksMNISTSpiNNakerEvent processingNeural networks have enabled great advances in recent times due mainly to improved parallel computing capabilities in accordance to Moore’s Law, which allowed reducing the time needed for the parameter learning of complex, multi-layered neural architectures. However, with silicon technology reaching its physical limits, new types of computing paradigms are needed to increase the power efficiency of learning algorithms, especially for dealing with deep spatio-temporal knowledge on embedded applications. With the goal of mimicking the brain’s power efficiency, new hardware architectures such as the SpiNNaker board have been built. Furthermore, recent works have shown that networks using spiking neurons as learning units can match classical neural networks in supervised tasks. In this paper, we show that the implementation of state-of-the-art models on both the MNIST and the event-based NMNIST digit recognition datasets is possible on neuromorphic hardware. We use two approaches, by directly converting a classical neural network to its spiking version and by training a spiking network from scratch. For both cases, software simulations and implementations into a SpiNNaker 103 machine were performed. Numerical results approaching the state of the art on digit recognition are presented, and a new method to decrease the spike rate needed for the task is proposed, which allows a significant reduction of the spikes (up to 34 times for a fully connected architecture) while preserving the accuracy of the system. With this method, we provide new insights on the capabilities offered by networks of spiking neurons to efficiently encode spatio-temporal information.Consejo Nacional de Ciencia Y Tecnología (México) FC2016-1961European Union's Horizon 2020 No 824164 HERMESMinisterio de Ciencia, Innovación y Universidades TEC2015-63884-C2-1-PElsevierArquitectura y Tecnología de ComputadoresConsejo Nacional de Ciencia y Tecnología (CONACYT). MéxicoEuropean Union (UE)Ministerio de Ciencia, Innovación y Universidades (MICINN). España2020info:eu-repo/semantics/articleinfo:eu-repo/semantics/submittedVersionapplication/pdfapplication/pdfhttps://hdl.handle.net/11441/102108https://doi.org/10.1016/j.neunet.2019.09.008reponame:idUS. Depósito de Investigación de la Universidad de Sevillainstname:Universidad de Sevilla (US)InglésNeural Networks, 121 (january 2020), 319-328.FC2016-1961824164 HERMESTEC2015-63884-C2-1-Phttps://www.sciencedirect.com/science/article/pii/S0893608019302692info:eu-repo/semantics/openAccessoai:idus.us.es:11441/1021082026-06-17T12:51:07Z
dc.title.none.fl_str_mv Event-driven implementation of deep spiking convolutional neural networks for supervised classification using the SpiNNaker neuromorphic platform
title Event-driven implementation of deep spiking convolutional neural networks for supervised classification using the SpiNNaker neuromorphic platform
spellingShingle Event-driven implementation of deep spiking convolutional neural networks for supervised classification using the SpiNNaker neuromorphic platform
Patiño Saucedo, Alberto
Neuromorphic hardware
Artificial neural networks
Spiking neural networks
MNIST
SpiNNaker
Event processing
title_short Event-driven implementation of deep spiking convolutional neural networks for supervised classification using the SpiNNaker neuromorphic platform
title_full Event-driven implementation of deep spiking convolutional neural networks for supervised classification using the SpiNNaker neuromorphic platform
title_fullStr Event-driven implementation of deep spiking convolutional neural networks for supervised classification using the SpiNNaker neuromorphic platform
title_full_unstemmed Event-driven implementation of deep spiking convolutional neural networks for supervised classification using the SpiNNaker neuromorphic platform
title_sort Event-driven implementation of deep spiking convolutional neural networks for supervised classification using the SpiNNaker neuromorphic platform
dc.creator.none.fl_str_mv Patiño Saucedo, Alberto
Rostro González, Horacio
Serrano Gotarredona, María Teresa
Linares Barranco, Bernabé
author Patiño Saucedo, Alberto
author_facet Patiño Saucedo, Alberto
Rostro González, Horacio
Serrano Gotarredona, María Teresa
Linares Barranco, Bernabé
author_role author
author2 Rostro González, Horacio
Serrano Gotarredona, María Teresa
Linares Barranco, Bernabé
author2_role author
author
author
dc.contributor.none.fl_str_mv Arquitectura y Tecnología de Computadores
Consejo Nacional de Ciencia y Tecnología (CONACYT). México
European Union (UE)
Ministerio de Ciencia, Innovación y Universidades (MICINN). España
dc.subject.none.fl_str_mv Neuromorphic hardware
Artificial neural networks
Spiking neural networks
MNIST
SpiNNaker
Event processing
topic Neuromorphic hardware
Artificial neural networks
Spiking neural networks
MNIST
SpiNNaker
Event processing
description Neural networks have enabled great advances in recent times due mainly to improved parallel computing capabilities in accordance to Moore’s Law, which allowed reducing the time needed for the parameter learning of complex, multi-layered neural architectures. However, with silicon technology reaching its physical limits, new types of computing paradigms are needed to increase the power efficiency of learning algorithms, especially for dealing with deep spatio-temporal knowledge on embedded applications. With the goal of mimicking the brain’s power efficiency, new hardware architectures such as the SpiNNaker board have been built. Furthermore, recent works have shown that networks using spiking neurons as learning units can match classical neural networks in supervised tasks. In this paper, we show that the implementation of state-of-the-art models on both the MNIST and the event-based NMNIST digit recognition datasets is possible on neuromorphic hardware. We use two approaches, by directly converting a classical neural network to its spiking version and by training a spiking network from scratch. For both cases, software simulations and implementations into a SpiNNaker 103 machine were performed. Numerical results approaching the state of the art on digit recognition are presented, and a new method to decrease the spike rate needed for the task is proposed, which allows a significant reduction of the spikes (up to 34 times for a fully connected architecture) while preserving the accuracy of the system. With this method, we provide new insights on the capabilities offered by networks of spiking neurons to efficiently encode spatio-temporal information.
publishDate 2020
dc.date.none.fl_str_mv 2020
dc.type.none.fl_str_mv info:eu-repo/semantics/article
info:eu-repo/semantics/submittedVersion
format article
status_str submittedVersion
dc.identifier.none.fl_str_mv https://hdl.handle.net/11441/102108
https://doi.org/10.1016/j.neunet.2019.09.008
url https://hdl.handle.net/11441/102108
https://doi.org/10.1016/j.neunet.2019.09.008
dc.language.none.fl_str_mv Inglés
language_invalid_str_mv Inglés
dc.relation.none.fl_str_mv Neural Networks, 121 (january 2020), 319-328.
FC2016-1961
824164 HERMES
TEC2015-63884-C2-1-P
https://www.sciencedirect.com/science/article/pii/S0893608019302692
dc.rights.none.fl_str_mv info:eu-repo/semantics/openAccess
eu_rights_str_mv openAccess
dc.format.none.fl_str_mv application/pdf
application/pdf
dc.publisher.none.fl_str_mv Elsevier
publisher.none.fl_str_mv Elsevier
dc.source.none.fl_str_mv reponame:idUS. Depósito de Investigación de la Universidad de Sevilla
instname:Universidad de Sevilla (US)
instname_str Universidad de Sevilla (US)
reponame_str idUS. Depósito de Investigación de la Universidad de Sevilla
collection idUS. Depósito de Investigación de la Universidad de Sevilla
repository.name.fl_str_mv
repository.mail.fl_str_mv
_version_ 1869409340253798400
score 15,198674

Registos relacionados