Supervised learning in spiking, neural networks with noise-threshold

Malu Zhang, Hong Qu*, Xiurui Xie, Juergen Kurths

*Corresponding author for this work

Research output: Contribution to journalArticle

19 Citations (Scopus)

Abstract

With a similar capability of processing spikes as biological neural systems, networks of spiking neurons are expected to achieve a performance similar to that of living brains. Despite the achievement of spiking neuron based applications, most of them assume noise-free condition for learning and testing. This assumption, though fairly general, ignores the fact that noise widely exists in spiking neural networks (SNNs) and the neural response can be significantly disturbed by noise. Therefore, how to deal with noise is an important issue in the applications of SNNs. Here, by analyzing strategies employed to make spiking neurons robust to noise, also inspired by biological neurons, we propose a strategy that train spiking neurons with a dynamic firing threshold named noise-threshold. The noise-threshold can be applied by the existing supervised learning methods to improve the noise tolerance of them. Experimental results show that, with a combination of noise-threshold, the anti-noise capability of the existing supervised learning methods improves significantly, and the trained neuron can precisely and reliably reproduce target sequences of spikes even under highly noisy conditions. More importantly, the SNNs-based computational model equipped with a noise-threshold is more robust and can achieve a good performance even with different types of noise. Therefore, the noise-threshold is significant to practical applications and theoretical researches of SNNs.

Original languageEnglish
Pages (from-to)333-349
Number of pages17
JournalNeurocomputing
Volume219
Early online date24 Sep 2016
DOIs
Publication statusPublished - 5 Jan 2017

Keywords

  • Spiking neurons
  • Noise-threshold
  • Supervised learning
  • Spiking neural networks (SNNs)
  • Anti-noise capability
  • RETINAL GANGLION-CELLS
  • IN-VIVO
  • NEURONS
  • PRECISION
  • INFORMATION
  • RELIABILITY
  • SIGNALS
  • TRAINS
  • MODEL

Cite this

Supervised learning in spiking, neural networks with noise-threshold. / Zhang, Malu; Qu, Hong; Xie, Xiurui; Kurths, Juergen.

In: Neurocomputing, Vol. 219, 05.01.2017, p. 333-349.

Research output: Contribution to journalArticle

Zhang, Malu ; Qu, Hong ; Xie, Xiurui ; Kurths, Juergen. / Supervised learning in spiking, neural networks with noise-threshold. In: Neurocomputing. 2017 ; Vol. 219. pp. 333-349.
@article{bf64f2e371df4f25862bedb925c48d0a,
title = "Supervised learning in spiking, neural networks with noise-threshold",
abstract = "With a similar capability of processing spikes as biological neural systems, networks of spiking neurons are expected to achieve a performance similar to that of living brains. Despite the achievement of spiking neuron based applications, most of them assume noise-free condition for learning and testing. This assumption, though fairly general, ignores the fact that noise widely exists in spiking neural networks (SNNs) and the neural response can be significantly disturbed by noise. Therefore, how to deal with noise is an important issue in the applications of SNNs. Here, by analyzing strategies employed to make spiking neurons robust to noise, also inspired by biological neurons, we propose a strategy that train spiking neurons with a dynamic firing threshold named noise-threshold. The noise-threshold can be applied by the existing supervised learning methods to improve the noise tolerance of them. Experimental results show that, with a combination of noise-threshold, the anti-noise capability of the existing supervised learning methods improves significantly, and the trained neuron can precisely and reliably reproduce target sequences of spikes even under highly noisy conditions. More importantly, the SNNs-based computational model equipped with a noise-threshold is more robust and can achieve a good performance even with different types of noise. Therefore, the noise-threshold is significant to practical applications and theoretical researches of SNNs.",
keywords = "Spiking neurons, Noise-threshold, Supervised learning, Spiking neural networks (SNNs), Anti-noise capability, RETINAL GANGLION-CELLS, IN-VIVO, NEURONS, PRECISION, INFORMATION, RELIABILITY, SIGNALS, TRAINS, MODEL",
author = "Malu Zhang and Hong Qu and Xiurui Xie and Juergen Kurths",
year = "2017",
month = "1",
day = "5",
doi = "10.1016/j.neucom.2016.09.044",
language = "English",
volume = "219",
pages = "333--349",
journal = "Neurocomputing",
issn = "0925-2312",
publisher = "ELSEVIER SCIENCE BV",

}

TY - JOUR

T1 - Supervised learning in spiking, neural networks with noise-threshold

AU - Zhang, Malu

AU - Qu, Hong

AU - Xie, Xiurui

AU - Kurths, Juergen

PY - 2017/1/5

Y1 - 2017/1/5

N2 - With a similar capability of processing spikes as biological neural systems, networks of spiking neurons are expected to achieve a performance similar to that of living brains. Despite the achievement of spiking neuron based applications, most of them assume noise-free condition for learning and testing. This assumption, though fairly general, ignores the fact that noise widely exists in spiking neural networks (SNNs) and the neural response can be significantly disturbed by noise. Therefore, how to deal with noise is an important issue in the applications of SNNs. Here, by analyzing strategies employed to make spiking neurons robust to noise, also inspired by biological neurons, we propose a strategy that train spiking neurons with a dynamic firing threshold named noise-threshold. The noise-threshold can be applied by the existing supervised learning methods to improve the noise tolerance of them. Experimental results show that, with a combination of noise-threshold, the anti-noise capability of the existing supervised learning methods improves significantly, and the trained neuron can precisely and reliably reproduce target sequences of spikes even under highly noisy conditions. More importantly, the SNNs-based computational model equipped with a noise-threshold is more robust and can achieve a good performance even with different types of noise. Therefore, the noise-threshold is significant to practical applications and theoretical researches of SNNs.

AB - With a similar capability of processing spikes as biological neural systems, networks of spiking neurons are expected to achieve a performance similar to that of living brains. Despite the achievement of spiking neuron based applications, most of them assume noise-free condition for learning and testing. This assumption, though fairly general, ignores the fact that noise widely exists in spiking neural networks (SNNs) and the neural response can be significantly disturbed by noise. Therefore, how to deal with noise is an important issue in the applications of SNNs. Here, by analyzing strategies employed to make spiking neurons robust to noise, also inspired by biological neurons, we propose a strategy that train spiking neurons with a dynamic firing threshold named noise-threshold. The noise-threshold can be applied by the existing supervised learning methods to improve the noise tolerance of them. Experimental results show that, with a combination of noise-threshold, the anti-noise capability of the existing supervised learning methods improves significantly, and the trained neuron can precisely and reliably reproduce target sequences of spikes even under highly noisy conditions. More importantly, the SNNs-based computational model equipped with a noise-threshold is more robust and can achieve a good performance even with different types of noise. Therefore, the noise-threshold is significant to practical applications and theoretical researches of SNNs.

KW - Spiking neurons

KW - Noise-threshold

KW - Supervised learning

KW - Spiking neural networks (SNNs)

KW - Anti-noise capability

KW - RETINAL GANGLION-CELLS

KW - IN-VIVO

KW - NEURONS

KW - PRECISION

KW - INFORMATION

KW - RELIABILITY

KW - SIGNALS

KW - TRAINS

KW - MODEL

U2 - 10.1016/j.neucom.2016.09.044

DO - 10.1016/j.neucom.2016.09.044

M3 - Article

VL - 219

SP - 333

EP - 349

JO - Neurocomputing

JF - Neurocomputing

SN - 0925-2312

ER -