NEURO-COMPUTATION TECHNIQUES IN SAMPLED-DATA ELECTROMAGNETIC-FIELD PROBLEMS

In this paper, a technique is introduced by which to extend the applic ability of the existing analytic solutions of electromagnetic field pr oblems to cases where random-noisy-sampled data (such as measurement o utputs) are available, rather than analytic input functions. We addres s those problems for which a theoretical solution exists in the form o f a superposition of some basis functions. The algorithm introduced em ploys this same set of basis functions, and finds the expansion coeffi cients by the use of an iterative error-minimization technique, which resembles those found in the process of training of artificial neural networks. In cases where the expansion functions are orthonormal, guar anteed fast convergence is proved. As well, we show how neuro-computat ion techniques can be employed to circumvent the effects of various ty pes of measurement errors and noise. Satisfactory performance of the a lgorithm is shown for a test problem driven by random inputs corrupted with various levels of Gaussian noise.