Gaussian Sum Filters for Recurrent Neural Networks training

Author

Todorovic, Branimir ; Stankovic, Miomir ; Moraga, Claudio

Author_Institution

Fac. of Occupational Safety, Nis Univ.

fYear

2006

fDate

25-27 Sept. 2006

Firstpage

53

Lastpage

57

Abstract

We consider the problem of recurrent neural network training as a Bayesian state estimation. The proposed algorithm uses Gaussian sum filter for nonlinear, non-Gaussian estimation of network outputs and synaptic weights. The performances of the proposed algorithm and other Bayesian filters are compared in noisy chaotic time series long-term prediction

Keywords

Bayes methods; Gaussian processes; learning (artificial intelligence); recurrent neural nets; time series; Bayesian state estimation; Gaussian sum filters; noisy chaotic time series; recurrent neural networks training; Bayesian methods; Chaos; Filters; Neural networks; Neurons; Nonlinear dynamical systems; Nonlinear equations; Probability density function; Recurrent neural networks; State estimation; Gaussian sum filter; Recurrent neural networks; divided difference filter; extended Kalman filter; sequential Bayesian estimation; unscented Kalman filter;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Network Applications in Electrical Engineering, 2006. NEUREL 2006. 8th Seminar on

Conference_Location

Belgrade, Serbia & Montenegro

Print_ISBN

1-4244-0433-9

Electronic_ISBN

1-4244-0433-9

Type

conf

DOI

10.1109/NEUREL.2006.341175

Filename

4147163