back to publications overview

Studies on Reservoir Initialization and Dynamics Shaping in Echo State Networks


Joschka Boedecker, Oliver Obst, Norbert Michael Mayer, and Minoru Asada. Studies on Reservoir Initialization and Dynamics Shaping in Echo State Networks. In Proceedings of the 17th European Symposium On Artificial Neural Networks (ESANN'09), pp. 227–232, D-Side Publications, Evere, Belgium, April 2009.


Download

[PDF] (132 kb)


Abstract

The fixed random connectivity of networks in reservoir computing leads to significant variation in performance. Only few problem specific optimization procedures are known to date.We study a general initialization method using permutation matrices and derive a new unsupervised learning rule based on intrinsic plasticity (IP) for echo state networks. Using three different benchmarks, we show that networks with permutation matrices for the reservoir connectivity have much longer memory than the other methods, but are also able to perform highly non-linear mappings. We also show that IP based on sigmoid transfer functions is limited concerning the output distributions that can be achieved.

BiBTeX Entry


@inproceedings{BOMA09,
	Address = {Evere, Belgium},
	Author = {Joschka Boedecker and Oliver Obst and Norbert Michael Mayer and Minoru Asada},
	Booktitle = {Proceedings of the 17th European Symposium On Artificial Neural Networks ({ESANN}'09)},
	Editor = {Michel Verleysen},
	Month = apr,
	Pages = {227--232},
	Publisher = {D-Side Publications},
	Title = {Studies on Reservoir Initialization and Dynamics Shaping in Echo State Networks},
	Year = 2009,
	Abstract = {The fixed random connectivity of networks in reservoir computing leads to significant variation in performance. 
Only few problem specific optimization procedures are known to date.
We study a general initialization method using permutation matrices and derive a new unsupervised learning 
rule based on intrinsic plasticity (IP) for echo state networks. Using three different benchmarks, we show that 
networks with permutation matrices for the reservoir connectivity have much longer memory than the other 
methods, but are also able to perform highly non-linear mappings. We also show that IP based on sigmoid 
transfer functions is limited concerning the output distributions that can be achieved.
},