A neural network approach to least squares estimation

The authors present a new neural network approach to the problem of least squares parameter estimation and identification in engineering applications. First, they define the fundamental estimation problems, which is reformulated into a form suitable for a neural network realization. After introducing the interconnected neural network architecture, the required inputs and the values of the connectivities among the processing elements are derived. A numerical example is presented to illustrate the detrimental effects of inevitable parameter variations and noise