|
|
Digital Library of the
European Council for Modelling and Simulation |
|
Title: |
Towards Evolutionary Deep Neural Networks |
|
Authors: |
Tomas H. Maul, Andrzej Bargiela, Siang-Yew Chong, Abdullahi
S. Adamu |
|
Published in:
|
(2014).ECMS 2014 Proceedings edited by:
Flaminio Squazzoni, Fabio
Baronio, Claudia Archetti,
Marco Castellani European Council for
Modeling and Simulation. doi:10.7148/2014 ISBN:
978-0-9564944-8-1 28th
European Conference on Modelling and Simulation, Brescia,
Italy, May 27th – 30th,
2014 |
|
Citation
format: |
Tomas
H. Maul, Andrzej Bargiela,
Siang-Yew Chong, Abdullahi S. Adamu
(2014). Towards Evolutionary Deep Neural Networks, ECMS
2014 Proceedings edited by: Flaminio Squazzoni, Fabio Baronio,
Claudia Archetti, Marco Castellani European
Council for Modeling and Simulation. doi:10.7148/2014-0319 |
|
DOI: |
http://dx.doi.org/10.7148/2014-0319 |
|
Abstract: |
This paper is
concerned with the problem of optimizing deep neural networks with diverse
transfer functions using evolutionary methods. Standard evolutionary (SEDeeNN) and cooperative coevolutionary
methods (CoDeeNN) were applied to three different
architectures characterized by different constraints on neural diversity. It
was found that (1) SEDeeNN (but not CoDeeNN) changes parameters uniformly across all layers,
(2) both evolutionary approaches can exhibit good convergence and
generalization properties, and (3) increased neural diversity improves both
convergence and generalization. In addition to clarifying the feasibility of
evolutionary deep neural networks, we suggests a
guiding principle for synergizing evolutionary and error gradient based
approaches through layerchange analysis. |
|
Full
text: |
|
