A HYBRID ALGORITHM TO SOLVE LARGE SCALE ELECTROMAGNETIC PROBLEMS
Abdelmadjid Nouicer - Mohamed Elhadi Latreche
In this paper, the use of feed forward neural networks (FNN) coupled with the wavelet transform to solve electromagnetic problems is investigated. The direct use of the FNN to solve large scale electromagnetic problems needs a lot of CPU time and computer memory because we deal with a large size of training data base. So, the wavelet transform is proposed in order to reduce the data base size, in other terms using wavelets coefficients as training data instead of the original signal. A simple example shows the feasibility of our approach.
Keywords: feed-forward neural networks, finite elements computation, large-scale problems, wavelet transforms