Face processing using one spike per neurone. Rullen, R. V., Gautrais, J., Delorme, A., & Thorpe, S. Biosystems, 48(1-3):229-239, Centre de Recherche Cerveau et Cognition, UMR 5549, Toulouse, France., Sep, 1998. abstract bibtex The speed with which neurones in the monkey temporal lobe can respond selectively to the presence of a face implies that processing may be possible using only one spike per neurone, a finding that is problematic for conventional rate coding models that need at least two spikes to estimate interspike interval. One way of avoiding this problem uses the fact that integrate-and-fire neurones will tend to fire at different times, with the most strongly activated neurones firing first (Thorpe, 1990, Parallel Processing in Neural Systems). Under such conditions, processing can be performed by using the order in which cells in a particular layer fire as a code. To test this idea, we have explored a range of architectures using SpikeNET (Thorpe and Gautrais, 1997, Neural Information Processing Systems, 9), a simulator designed for modelling large populations of integrate-and-fire neurones. One such network used a simple four-layer feed-forward architecture to detect and localise the presence of human faces in natural images. Performance of the model was tested with a large range of grey-scale images of faces and other objects and was found to be remarkably good by comparison with more classic image processing techniques. The most remarkable feature of these results is that they were obtained using a purely feed-forward neural network in which none of the neurones fired more than one spike (thus ruling out conventional rate coding mechanisms). It thus appears that the combination of asynchronous spike propagation and rank order coding may provide an important key to understanding how the nervous system can achieve such a huge amount of processing in so little time.
@article{ VanRullen_etal98,
author = {R. Van Rullen and J. Gautrais and A. Delorme and S. Thorpe},
title = {Face processing using one spike per neurone},
journal = {Biosystems},
year = {1998},
volume = {48},
pages = {229-239},
number = {1-3},
month = {Sep},
abstract = {The speed with which neurones in the monkey temporal lobe can respond
selectively to the presence of a face implies that processing may
be possible using only one spike per neurone, a finding that is problematic
for conventional rate coding models that need at least two spikes
to estimate interspike interval. One way of avoiding this problem
uses the fact that integrate-and-fire neurones will tend to fire
at different times, with the most strongly activated neurones firing
first (Thorpe, 1990, Parallel Processing in Neural Systems). Under
such conditions, processing can be performed by using the order in
which cells in a particular layer fire as a code. To test this idea,
we have explored a range of architectures using SpikeNET (Thorpe
and Gautrais, 1997, Neural Information Processing Systems, 9), a
simulator designed for modelling large populations of integrate-and-fire
neurones. One such network used a simple four-layer feed-forward
architecture to detect and localise the presence of human faces in
natural images. Performance of the model was tested with a large
range of grey-scale images of faces and other objects and was found
to be remarkably good by comparison with more classic image processing
techniques. The most remarkable feature of these results is that
they were obtained using a purely feed-forward neural network in
which none of the neurones fired more than one spike (thus ruling
out conventional rate coding mechanisms). It thus appears that the
combination of asynchronous spike propagation and rank order coding
may provide an important key to understanding how the nervous system
can achieve such a huge amount of processing in so little time.},
address = {Centre de Recherche Cerveau et Cognition, UMR 5549, Toulouse, France.},
keywords = {*Action Potentials | *Face | Female | Human | Male | Models, Neurological
| Neurons/*physiology | *Visual Perception | 1999/01/14 00:01}
}
Downloads: 0
{"_id":"ExWp6MFtomE2gPSGS","bibbaseid":"rullen-gautrais-delorme-thorpe-faceprocessingusingonespikeperneurone-1998","downloads":0,"creationDate":"2015-02-08T05:14:56.526Z","title":"Face processing using one spike per neurone","author_short":["Rullen, R.<nbsp>V.","Gautrais, J.","Delorme, A.","Thorpe, S."],"year":1998,"bibtype":"article","biburl":"http://cnslab.mb.jhu.edu/niebase.bib","bibdata":{"abstract":"The speed with which neurones in the monkey temporal lobe can respond selectively to the presence of a face implies that processing may be possible using only one spike per neurone, a finding that is problematic for conventional rate coding models that need at least two spikes to estimate interspike interval. One way of avoiding this problem uses the fact that integrate-and-fire neurones will tend to fire at different times, with the most strongly activated neurones firing first (Thorpe, 1990, Parallel Processing in Neural Systems). Under such conditions, processing can be performed by using the order in which cells in a particular layer fire as a code. To test this idea, we have explored a range of architectures using SpikeNET (Thorpe and Gautrais, 1997, Neural Information Processing Systems, 9), a simulator designed for modelling large populations of integrate-and-fire neurones. One such network used a simple four-layer feed-forward architecture to detect and localise the presence of human faces in natural images. Performance of the model was tested with a large range of grey-scale images of faces and other objects and was found to be remarkably good by comparison with more classic image processing techniques. The most remarkable feature of these results is that they were obtained using a purely feed-forward neural network in which none of the neurones fired more than one spike (thus ruling out conventional rate coding mechanisms). It thus appears that the combination of asynchronous spike propagation and rank order coding may provide an important key to understanding how the nervous system can achieve such a huge amount of processing in so little time.","address":"Centre de Recherche Cerveau et Cognition, UMR 5549, Toulouse, France.","author":["Rullen, R. Van","Gautrais, J.","Delorme, A.","Thorpe, S."],"author_short":["Rullen, R.<nbsp>V.","Gautrais, J.","Delorme, A.","Thorpe, S."],"bibtex":"@article{ VanRullen_etal98,\n author = {R. Van Rullen and J. Gautrais and A. Delorme and S. Thorpe},\n title = {Face processing using one spike per neurone},\n journal = {Biosystems},\n year = {1998},\n volume = {48},\n pages = {229-239},\n number = {1-3},\n month = {Sep},\n abstract = {The speed with which neurones in the monkey temporal lobe can respond\n\tselectively to the presence of a face implies that processing may\n\tbe possible using only one spike per neurone, a finding that is problematic\n\tfor conventional rate coding models that need at least two spikes\n\tto estimate interspike interval. One way of avoiding this problem\n\tuses the fact that integrate-and-fire neurones will tend to fire\n\tat different times, with the most strongly activated neurones firing\n\tfirst (Thorpe, 1990, Parallel Processing in Neural Systems). Under\n\tsuch conditions, processing can be performed by using the order in\n\twhich cells in a particular layer fire as a code. To test this idea,\n\twe have explored a range of architectures using SpikeNET (Thorpe\n\tand Gautrais, 1997, Neural Information Processing Systems, 9), a\n\tsimulator designed for modelling large populations of integrate-and-fire\n\tneurones. One such network used a simple four-layer feed-forward\n\tarchitecture to detect and localise the presence of human faces in\n\tnatural images. Performance of the model was tested with a large\n\trange of grey-scale images of faces and other objects and was found\n\tto be remarkably good by comparison with more classic image processing\n\ttechniques. The most remarkable feature of these results is that\n\tthey were obtained using a purely feed-forward neural network in\n\twhich none of the neurones fired more than one spike (thus ruling\n\tout conventional rate coding mechanisms). It thus appears that the\n\tcombination of asynchronous spike propagation and rank order coding\n\tmay provide an important key to understanding how the nervous system\n\tcan achieve such a huge amount of processing in so little time.},\n address = {Centre de Recherche Cerveau et Cognition, UMR 5549, Toulouse, France.},\n keywords = {*Action Potentials | *Face | Female | Human | Male | Models, Neurological\n\t| Neurons/*physiology | *Visual Perception | 1999/01/14 00:01}\n}","bibtype":"article","id":"VanRullen_etal98","journal":"Biosystems","key":"VanRullen_etal98","keywords":"*Action Potentials | *Face | Female | Human | Male | Models, Neurological | Neurons/*physiology | *Visual Perception | 1999/01/14 00:01","month":"Sep","number":"1-3","pages":"229-239","title":"Face processing using one spike per neurone","type":"article","volume":"48","year":"1998","bibbaseid":"rullen-gautrais-delorme-thorpe-faceprocessingusingonespikeperneurone-1998","role":"author","urls":{},"keyword":["*Action Potentials | *Face | Female | Human | Male | Models","Neurological | Neurons/*physiology | *Visual Perception | 1999/01/14 00:01"],"downloads":0},"search_terms":["face","processing","using","one","spike","per","neurone","rullen","gautrais","delorme","thorpe"],"keywords":["*action potentials | *face | female | human | male | models","neurological | neurons/*physiology | *visual perception | 1999/01/14 00:01"],"authorIDs":[],"dataSources":["ErLXoH8mqSjESnrN5"]}