Anti Boundary Effect Wavelet Decomposition Echo State Networks

We propose a novel approach based on wavelet decomposition and echo state networks to discover the multiscale dynamics of time series which we call anti-boundary-effect wavelet decomposition and echo state networks (ABE-WESNs). ABE-WESNs use the wavelet decomposition as preprocessing steps and choose a matched ESNs for every scale level. We use the data extension methods to overcome the boundary effect. The introduced weight factors can both resolve the problem of cumulation of errors resulting from the wavelet decomposition. Experiments and engineering applications show that the ABE-WESNs can accurately model and predict some time series with multiscale properties.

[1]  Peng Xi-yuan,et al.  Researches on Time Series Prediction with Echo State Networks , 2010 .

[2]  Helmut Hauser,et al.  Echo state networks with filter neurons and a delay&sum readout , 2010, Neural Networks.

[3]  Yu Peng,et al.  Mobile communication traffic forecast based on a new fuzzy model , 2009, 2009 IEEE Instrumentation and Measurement Technology Conference.

[4]  Francis wyffels,et al.  Using reservoir computing in a decomposition approach for time series prediction , 2008 .

[5]  Benjamin Schrauwen,et al.  Band-pass Reservoir Computing , 2008, 2008 IEEE International Joint Conference on Neural Networks (IEEE World Congress on Computational Intelligence).

[6]  Herbert Jaeger,et al.  Reservoir computing approaches to recurrent neural network training , 2009, Comput. Sci. Rev..

[7]  Mark J. Shensa,et al.  The discrete wavelet transform: wedding the a trous and Mallat algorithms , 1992, IEEE Trans. Signal Process..

[8]  Simon Haykin,et al.  Decoupled echo state networks with lateral inhibition , 2007, Neural Networks.

[9]  Herbert Jaeger,et al.  Discovering multiscale dynamical features with hierarchical Echo State Networks , 2008 .