Learning short synfire chains by self-organization
Learning short synfire chains by self-organization
We develop a model of cortical coding of stimuli by the sequences of activation patterns that they ignite in an initially random network. Hebbian learning then stabilizes these sequences, making them attractors of the dynamics. There is a competition between the capacity of the network and the stability of the sequences; for small stability parameter epsilon (the strength of the mean stabilizing PSP in the neurons in a learned sequence) the capacity is proportional to 1/epsilon2. For epsilon of the order of or less than the PSPs of the untrained network, the capacity exceeds that for sequences learned from tabula rasa.
357-363
Hertz, John
24c0aff2-8219-4099-a77d-c3d220242368
Prügel-Bennett, Adam
b107a151-1751-4d8b-b8db-2c395ac4e14e
1996
Hertz, John
24c0aff2-8219-4099-a77d-c3d220242368
Prügel-Bennett, Adam
b107a151-1751-4d8b-b8db-2c395ac4e14e
Hertz, John and Prügel-Bennett, Adam
(1996)
Learning short synfire chains by self-organization.
Network: Computatioon in Neural Systems, 7 (2), .
Abstract
We develop a model of cortical coding of stimuli by the sequences of activation patterns that they ignite in an initially random network. Hebbian learning then stabilizes these sequences, making them attractors of the dynamics. There is a competition between the capacity of the network and the stability of the sequences; for small stability parameter epsilon (the strength of the mean stabilizing PSP in the neurons in a learned sequence) the capacity is proportional to 1/epsilon2. For epsilon of the order of or less than the PSPs of the untrained network, the capacity exceeds that for sequences learned from tabula rasa.
Text
MO3A5F20082D73967.pdf
- Other
More information
Published date: 1996
Organisations:
Southampton Wireless Group
Identifiers
Local EPrints ID: 254243
URI: http://eprints.soton.ac.uk/id/eprint/254243
PURE UUID: b59e1bf1-07cf-454d-882a-538f873d8572
Catalogue record
Date deposited: 12 Jan 2001
Last modified: 14 Mar 2024 05:32
Export record
Contributors
Author:
John Hertz
Author:
Adam Prügel-Bennett
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics