论文信息 - The Application of modified ESN in chaotic time series prediction

The Application of modified ESN in chaotic time series prediction

The parameters selection of ESN (Echo State Network) is excessively dependent on human experience, it is difficult to produce the corresponding optimal parameters for specific problem, resulting in severely restricted in practice. In view of this, a chaotic time series prediction model is proposed in this paper, and the model is based on differential evolution algorithm and the echo state network. With this model, training the input sample sequence to find the network's parameters which is suitable for the data characteristics at first, then use the ideal parameters to predict chaotic time series. In the prediction of the typical chaotic time series generated by Lorenz system, this method can establish a suitable echo state network based on the data characteristics effectively, and gets satisfactory results.

Deming Liu | Yong Zhang | Yongbing Yu

[1] Yongji Wang,et al. Modeling of Cortical Signals Using Optimized Echo State Networks with Leaky Integrator Neurons , 2009, ICONIP.

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

[3] Teresa Bernarda Ludermir,et al. Genetic algorithm for reservoir computing optimization , 2009, 2009 International Joint Conference on Neural Networks.

[4] Rainer Storn,et al. Differential Evolution – A Simple and Efficient Heuristic for global Optimization over Continuous Spaces , 1997, J. Glob. Optim..

[5] Herbert Jaeger,et al. The''echo state''approach to analysing and training recurrent neural networks , 2001 .

[6] Amit Konar,et al. Two-Dimensional IIR Filter Design with Modern Search Heuristics: a Comparative Study , 2006, Int. J. Comput. Intell. Appl..