TY - GEN
T1 - Recurrent neural networks training with stable risk-sensitive Kalman filter algorithm
AU - Yu, Wen
AU - De Jesús Rubio, José
AU - Li, Xiaoou
PY - 2005
Y1 - 2005
N2 - Compared to normal learning algorithms, for example backpropagation, Kalman filter-based algorithm has some better properties, such as faster convergence. In this paper, Kalman filter is modified with a risk-sensitive cost criterion, we call it as risk-sensitive Kalman filter. This new algorithm is applied to train recurrent neural networks for nonlinear system identification. Input-to-state stability is used to prove that the risk-sensitive Kalman filter training is stable. The contributions of this paper are: 1) the risk-sensitive Kalman filter is used for the state-space recurrent neural networks training, 2) the stability of the risk-sensitive Kalman filter is proved.
AB - Compared to normal learning algorithms, for example backpropagation, Kalman filter-based algorithm has some better properties, such as faster convergence. In this paper, Kalman filter is modified with a risk-sensitive cost criterion, we call it as risk-sensitive Kalman filter. This new algorithm is applied to train recurrent neural networks for nonlinear system identification. Input-to-state stability is used to prove that the risk-sensitive Kalman filter training is stable. The contributions of this paper are: 1) the risk-sensitive Kalman filter is used for the state-space recurrent neural networks training, 2) the stability of the risk-sensitive Kalman filter is proved.
UR - http://www.scopus.com/inward/record.url?scp=33745965821&partnerID=8YFLogxK
U2 - 10.1109/IJCNN.2005.1555937
DO - 10.1109/IJCNN.2005.1555937
M3 - Contribución a la conferencia
AN - SCOPUS:33745965821
SN - 0780390482
SN - 9780780390485
T3 - Proceedings of the International Joint Conference on Neural Networks
SP - 700
EP - 705
BT - Proceedings of the International Joint Conference on Neural Networks, IJCNN 2005
T2 - International Joint Conference on Neural Networks, IJCNN 2005
Y2 - 31 July 2005 through 4 August 2005
ER -