Accurate and efficient representations of exact electromagnetic scattering models using artificial neural networks

An efficient method to accurately model electromagnetic scattering based on artificial neural networks (ANNs) is described. Adaptive ANNs are utilised as approximators to model reflection from a lossless dielectric slab and scattering from a sphere. Correspondence to Mie scattering is achieved to within a normalised RMS error of 1.9%. Applications to the modelling of scattering from particle distributions are discussed