Online learning in Bayesian spiking neurons

Levin Kuhlmann; Michael Hauser-Raspe; Jonathan Manton; David B. Grayden; Jonathan Tapson; André van Schaik

doi:10.1109/IJCNN.2012.6252370

Online learning in Bayesian spiking neurons

Levin Kuhlmann, Michael Hauser-Raspe, Jonathan Manton, David B. Grayden, Jonathan Tapson, André van Schaik

MARCS Institute for Brain, Behaviour & Development

Research output: Chapter in Book / Conference Paper › Conference Paper › peer-review

1 Citation (Scopus)

Abstract

Bayesian Spiking Neurons (BSNs) provide a probabilistic interpretation of how neurons can perform inference and learning. Learning in a single BSN can be formulated as an online maximum-likelihood expectation-maximisation (ML-EM) algorithm. This form of learning is quite slow. Here, an alternative to this learning algorithm, called Fast Learning (FL), is presented. The FL algorithm is shown to have acceptable convergence performance when compared to the ML-EM algorithm. Moreover, for our implementations the FL algorithm is approximately 25 times faster than the ML-EM algorithm. Although only approximate, the FL algorithm therefore makes learning in hierarchical BSN networks much more tractable.

Original language	English
Title of host publication	Proceedings of the 2012 International Joint Conference on Neural Networks (IJCNN): 10-15 June 2012, Brisbane, Qld.
Publisher	IEEE
Pages	1-6
Number of pages	6
ISBN (Print)	9781467314909
DOIs	https://doi.org/10.1109/IJCNN.2012.6252370
Publication status	Published - 2012
Event	International Joint Conference on Neural Networks - Duration: 10 Jun 2012 → …

Conference

Conference	International Joint Conference on Neural Networks
Period	10/06/12 → …

Keywords

Bayesian spiking neurons
algorithms
learning
mathematical models
neurons
online learning

Access to Document

10.1109/IJCNN.2012.6252370

Cite this

@inproceedings{3a96c8b30a2c4f91a3fa5a9c6728e0a7,

title = "Online learning in Bayesian spiking neurons",

abstract = "Bayesian Spiking Neurons (BSNs) provide a probabilistic interpretation of how neurons can perform inference and learning. Learning in a single BSN can be formulated as an online maximum-likelihood expectation-maximisation (ML-EM) algorithm. This form of learning is quite slow. Here, an alternative to this learning algorithm, called Fast Learning (FL), is presented. The FL algorithm is shown to have acceptable convergence performance when compared to the ML-EM algorithm. Moreover, for our implementations the FL algorithm is approximately 25 times faster than the ML-EM algorithm. Although only approximate, the FL algorithm therefore makes learning in hierarchical BSN networks much more tractable.",

keywords = "Bayesian spiking neurons, algorithms, learning, mathematical models, neurons, online learning",

author = "Levin Kuhlmann and Michael Hauser-Raspe and Jonathan Manton and Grayden, {David B.} and Jonathan Tapson and Schaik, {Andr{\'e} van}",

year = "2012",

doi = "10.1109/IJCNN.2012.6252370",

language = "English",

isbn = "9781467314909",

pages = "1--6",

booktitle = "Proceedings of the 2012 International Joint Conference on Neural Networks (IJCNN): 10-15 June 2012, Brisbane, Qld.",

publisher = "IEEE",

note = "International Joint Conference on Neural Networks ; Conference date: 10-06-2012",

}

TY - GEN

T1 - Online learning in Bayesian spiking neurons

AU - Kuhlmann, Levin

AU - Hauser-Raspe, Michael

AU - Manton, Jonathan

AU - Grayden, David B.

AU - Tapson, Jonathan

AU - Schaik, André van

PY - 2012

Y1 - 2012

N2 - Bayesian Spiking Neurons (BSNs) provide a probabilistic interpretation of how neurons can perform inference and learning. Learning in a single BSN can be formulated as an online maximum-likelihood expectation-maximisation (ML-EM) algorithm. This form of learning is quite slow. Here, an alternative to this learning algorithm, called Fast Learning (FL), is presented. The FL algorithm is shown to have acceptable convergence performance when compared to the ML-EM algorithm. Moreover, for our implementations the FL algorithm is approximately 25 times faster than the ML-EM algorithm. Although only approximate, the FL algorithm therefore makes learning in hierarchical BSN networks much more tractable.

AB - Bayesian Spiking Neurons (BSNs) provide a probabilistic interpretation of how neurons can perform inference and learning. Learning in a single BSN can be formulated as an online maximum-likelihood expectation-maximisation (ML-EM) algorithm. This form of learning is quite slow. Here, an alternative to this learning algorithm, called Fast Learning (FL), is presented. The FL algorithm is shown to have acceptable convergence performance when compared to the ML-EM algorithm. Moreover, for our implementations the FL algorithm is approximately 25 times faster than the ML-EM algorithm. Although only approximate, the FL algorithm therefore makes learning in hierarchical BSN networks much more tractable.

KW - Bayesian spiking neurons

KW - algorithms

KW - learning

KW - mathematical models

KW - neurons

KW - online learning

UR - http://handle.uws.edu.au:8081/1959.7/521291

UR - http://www.ieee-wcci2012.org/ieee-wcci2012/images/stories/wcci/cfp/ijcnn122.pdf

U2 - 10.1109/IJCNN.2012.6252370

DO - 10.1109/IJCNN.2012.6252370

M3 - Conference Paper

SN - 9781467314909

SP - 1

EP - 6

BT - Proceedings of the 2012 International Joint Conference on Neural Networks (IJCNN): 10-15 June 2012, Brisbane, Qld.

PB - IEEE

T2 - International Joint Conference on Neural Networks

Y2 - 10 June 2012

ER -

Online learning in Bayesian spiking neurons

Abstract

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this