EINSTEINIAN NEURAL-NETWORK FOR SPECTRUM ESTIMATION

A model-based neural network is developed for spectrum estimation. Its architecture and learning mechanism are founded on the Einsteinian in terpretation of the spectrum as a probability distribution ofphotons. By considering a spectrum as an ensemble of photons, we derive the neu ral learning mechanism from the basic physical principle of entropy ma ximization of a canonical ensemble. This neural network is applied to characterizing a recently observed phenomenon known as equatorial iono spheric clutter that significantly affects operations of over-the-hori zon (OTH) radars and communication links using high frequency radiowav es propagating through the ionosphere. We utilize a specific parameter ization of the internal spectral model, which is derived from the phys ical principles of the propagation of electromagnetic waves through a turbulent ionosphere. A set of parameters characterizing equatorial io nospheric clutter is estimated The developed technique may have a broa d applicability in scientific data analysis. (C) 1997 Elsevier Science Ltd. All rights reserved.