Optimization and applications of echo state networks with leaky-integrator neurons
- PMID: 17517495
- DOI: 10.1016/j.neunet.2007.04.016
Optimization and applications of echo state networks with leaky-integrator neurons
Abstract
Standard echo state networks (ESNs) are built from simple additive units with a sigmoid activation function. Here we investigate ESNs whose reservoir units are leaky integrator units. Units of this type have individual state dynamics, which can be exploited in various ways to accommodate the network to the temporal characteristics of a learning task. We present stability conditions, introduce and investigate a stochastic gradient descent method for the optimization of the global learning parameters (input and output feedback scalings, leaking rate, spectral radius) and demonstrate the usefulness of leaky-integrator ESNs for (i) learning very slow dynamic systems and replaying the learnt system at different speeds, (ii) classifying relatively slow and noisy time series (the Japanese Vowel dataset--here we obtain a zero test error rate), and (iii) recognizing strongly time-warped dynamic patterns.
Similar articles
-
Re-visiting the echo state property.Neural Netw. 2012 Nov;35:1-9. doi: 10.1016/j.neunet.2012.07.005. Epub 2012 Jul 23. Neural Netw. 2012. PMID: 22885243
-
Effects of spectral radius and settling time in the performance of echo state networks.Neural Netw. 2009 Sep;22(7):861-3. doi: 10.1016/j.neunet.2009.03.021. Epub 2009 Apr 23. Neural Netw. 2009. PMID: 19423285
-
Learning grammatical structure with Echo State Networks.Neural Netw. 2007 Apr;20(3):424-32. doi: 10.1016/j.neunet.2007.04.013. Epub 2007 May 3. Neural Netw. 2007. PMID: 17556116
-
Non-linear dynamics in neural networks.Prog Brain Res. 1994;102:371-82. doi: 10.1016/S0079-6123(08)60553-1. Prog Brain Res. 1994. PMID: 7800827 Review.
-
Grid cells: the position code, neural network models of activity, and the problem of learning.Hippocampus. 2008;18(12):1283-300. doi: 10.1002/hipo.20519. Hippocampus. 2008. PMID: 19021263 Review.
Cited by
-
Heave compensation prediction based on echo state network with correntropy induced loss function.PLoS One. 2019 Jun 13;14(6):e0217361. doi: 10.1371/journal.pone.0217361. eCollection 2019. PLoS One. 2019. PMID: 31194791 Free PMC article.
-
Multiscale model of an inhibitory network shows optimal properties near bifurcation.Phys Rev Lett. 2011 Jun 10;106(23):238109. doi: 10.1103/PhysRevLett.106.238109. Epub 2011 Jun 10. Phys Rev Lett. 2011. PMID: 21770552 Free PMC article.
-
Optimal nonlinear information processing capacity in delay-based reservoir computers.Sci Rep. 2015 Sep 11;5:12858. doi: 10.1038/srep12858. Sci Rep. 2015. PMID: 26358528 Free PMC article.
-
Echo state networks for the recognition of type 1 Brugada syndrome from conventional 12-LEAD ECG.Heliyon. 2024 Feb 1;10(3):e25404. doi: 10.1016/j.heliyon.2024.e25404. eCollection 2024 Feb 15. Heliyon. 2024. PMID: 38333823 Free PMC article.
-
Optoelectronic reservoir computing.Sci Rep. 2012;2:287. doi: 10.1038/srep00287. Epub 2012 Feb 27. Sci Rep. 2012. PMID: 22371825 Free PMC article.
Publication types
MeSH terms
LinkOut - more resources
Full Text Sources
Other Literature Sources