Neural network model for solving integral equation of acoustic scattering using wavelet basis

In this work, neural networks (NNs) are trained in order to obtain a fast and efficient solution of the integral equation of acoustic scattering. Wavelets sparsification methods are utilized to reduce the NN size and complexity. The non-uniqueness problem which arises in solving this integral equation at characteristic frequencies can, also, be solved using such network without any additional computational load. Experimental results show excellent agreement between the NN-based solution and the analytical solution of a spherical scatterer. Copyright (C) 2006 John Wiley & Sons, Ltd.