论文信息 - Generalized Backpropagation through Time for Continuous Time Neural Networks and Discrete Time Measurements

Generalized Backpropagation through Time for Continuous Time Neural Networks and Discrete Time Measurements

This paper deals with the problem of identification of continuous time dynamic neural networks when the measurements are given only at discrete time moments, not necessarily uniformly distributed. It is shown that the modified adjoint system, generating the gradient of the performance index, is a continuous-time system with jumps of state variables at moments corresponding to moments of measurements.

Krzysztof Fujarewicz | Adam Galuszka | A. Gałuszka | K. Fujarewicz

[1] Eric A. Wan,et al. Diagrammatic Derivation of Gradient Algorithms for Neural Networks , 1996, Neural Computation.

[2] Ronald J. Williams,et al. A Learning Algorithm for Continually Running Fully Recurrent Neural Networks , 1989, Neural Computation.

[3] Barak A. Pearlmutter. Gradient calculations for dynamic recurrent neural networks: a survey , 1995, IEEE Trans. Neural Networks.

[4] K. Fujarewicz. Identification and suboptimal control of heat exchanger using generalized back propagation through time , 2000 .

[5] Pierre Baldi,et al. Gradient descent learning algorithm overview: a general dynamical systems perspective , 1995, IEEE Trans. Neural Networks.

[6] Thomas Kailath,et al. Linear Systems , 1980 .

[7] Paul J. Werbos,et al. Backpropagation Through Time: What It Does and How to Do It , 1990, Proc. IEEE.

[8] Kumpati S. Narendra,et al. Gradient methods for the optimization of dynamical systems containing neural networks , 1991, IEEE Trans. Neural Networks.

[9] Krzysztof Fujarewicz,et al. Evaluation of a Helicopter Model Using Generalized Back Propagation Through Time , 2003 .

[10] M. E. Ahmed. Block partial derivative and its application to neural-net-based direct model-reference adaptive control , 1994 .