%0 Report %1 TR229 %A Heister, Frank %A Müller, Rainer %C Department of Computer Science %D 1999 %K myown %N 229 %T An approach for the identification of nonlinear, dynamic processes with Kalman Filter-trained recurrent neural structures. %U http://www-info3.informatik.uni-wuerzburg.de/TR/tr229.pdf %X In this article we demonstrate the identification of a nonlinear, dynamic process with recurrent neural structures. The employed network-structure is a Recurrent Multilayer Perceptron (RMLP), which combines feedforward-- and recurrent architectures. We will show that RMLPs are capable of learning the temporal behavior and characteristic of an arbitrary, nonlinear, dynamic process. Apart from conventional gradient-based algorithms, a sophisticated statistical method has been considered for this challenging task - Global Extended Kalman Filtering (GEKF). This powerful algorithm yields neural structures with a significantly better performance, compared to conventional gradient-based approaches. The new element in this work is the application of the GEKF-Algorithm for recurrent neural structures, which are employed in the identification of nonlinear, dynamic processes. In order to supervise the quality of network-training, appropriate performance-indexes for neural identification are introduced. The distribution of the Moving Average Squared Error (MASE) is employed as an objective optimality-criterion, in order to survey the actual performance of recurrent neural structures during training.

@techreport{TR229, abstract = { In this article we demonstrate the identification of a nonlinear, dynamic process with recurrent neural structures. The employed network-structure is a Recurrent Multilayer Perceptron (RMLP), which combines feedforward-- and recurrent architectures. We will show that RMLPs are capable of learning the temporal behavior and characteristic of an arbitrary, nonlinear, dynamic process. Apart from conventional gradient-based algorithms, a sophisticated statistical method has been considered for this challenging task - Global Extended Kalman Filtering (GEKF). This powerful algorithm yields neural structures with a significantly better performance, compared to conventional gradient-based approaches. The new element in this work is the application of the GEKF-Algorithm for recurrent neural structures, which are employed in the identification of nonlinear, dynamic processes. In order to supervise the quality of network-training, appropriate performance-indexes for neural identification are introduced. The distribution of the Moving Average Squared Error (MASE) is employed as an objective optimality-criterion, in order to survey the actual performance of recurrent neural structures during training. }, added-at = {2015-06-18T10:00:28.000+0200}, address = {Department of Computer Science}, author = {Heister, Frank and Müller, Rainer}, biburl = {https://www.bibsonomy.org/bibtex/247d0f4bf4e55c3f578f8fb5223e0ae30/trcsuniwue}, interhash = {693dd17e0c047cadacb8a3430b664d14}, intrahash = {47d0f4bf4e55c3f578f8fb5223e0ae30}, keywords = {myown}, month = may, number = 229, timestamp = {2015-06-18T10:00:28.000+0200}, title = {An approach for the identification of nonlinear, dynamic processes with Kalman Filter-trained recurrent neural structures.}, type = {Technical Report}, url = {http://www-info3.informatik.uni-wuerzburg.de/TR/tr229.pdf}, year = 1999 }