Processing Symbolic Sequences by Recurrent Neural Networks Trained by Kalman Filter-Based Algorithms

Abstract

Kalman filter (KF)-based techniques used for recurrent neural networks (RNNs) training on real-valued time series have already shown their potential. On the other hand gradient descent approaches such as back-propagation through time (BPTT) or real-time recurrent learning (RTRL) algorithms are still widely used by researchers working with symbolic sequences… (More)

Topics

3 Figures and Tables

Cite this paper

@inproceedings{Cernansk2006ProcessingSS, title={Processing Symbolic Sequences by Recurrent Neural Networks Trained by Kalman Filter-Based Algorithms}, author={Michal Cernansk{\'y} and Matej Makula and Lubica Benuskov{\'a}}, year={2006} }