Complex Optimization of a Recurrent Neural Network

Shows how to use the complex method to optimize a black-box neural network model of a load-sensing h

This package includes files for modelling nonlinear dynamic systems using a recurrent generalized neural network. The learning scheme uses the complex method of nonlinear nonderivative optimization, thereby avoiding the problems of computing the derivative of an infinite impulse response filter such as a recurrent neural network.

This package includes files for modelling nonlinear dynamic systems using a recurrent generalized neural network. The learning scheme uses the complex method of nonlinear nonderivative optimization, thereby avoiding the problems of computing the derivative of an infinite impulse response filter such as a recurrent neural network.

The example given is the modelling of a load-sensing hydraulic pump. The model output is the pump flow, as a response to inputs of pump pressure and the pressure in the control piston. Real experimental data is included.

For further details, refer to:
T. Wiens, R. Burton, G. Schoenau, D. Bitner, "Recursive Generalized Neural Networks (RGNN) for the Modeling of a Load Sensing Pump," Bath Symposium on Power Transmission and Motion Control, Sept 2008.
http://homepage.usask.ca/~tkw954/
http://blog.nutaksas.com

Note that this package requires the "Complex Method of Optimization" package in your path, available from http://www.mathworks.com/matlabcentral/fileexchange/loadFile.do?objectId=18342&objectType=FILE