BibTeX
@MISC{Brigham09self-organisedlearning,
author = {Marco Brigham},
title = {Self-organised learning in the . . .},
year = {2009}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
A review of the Chialvo-Bak model is presented, for the two-layer neural network topology. A novel Markov Chain representation is proposed that yields several important analytical quantities and supports a learning convergence argument. The power law regime is re-examined under this new representation and is found to be limited to learning under small mapping changes. A parallel between the power law regime and the biological neural avalanches is proposed. A mechanism to avoid the permanent tagging of synaptic weights of the selective punishment rule is proposed. i Acknowledgements I wish to thank Dr. Mark van Rossum for his tireless support and attentive guidance, and for having accepted to supervise me in the first place. To Dr. J. Michael Herrmann I wish to thank the very creative and rewarding discussions on the holistic merits of the Chialvo-Bak model.
Keyphrases
power law regime chialvo-bak model michael herrmann novel markov chain representation permanent tagging precious feedback rewarding discussion biological neural avalanche attentive guidance mark van rossum small mapping change wolfgang maass new representation first place two-layer neural network topology several important analytical quantity tireless support fruitful discussion holistic merit learning convergence argument t.u. graz synaptic weight selective punishment rule theoretical computer science
