The Method of Artificial Neural Networks in Retrieving Vertical Profiles of Atmospheric Parameters

摘要

The method of artificial neural networks (ANNs) and the method of principal components are considered jointly as applied to remote sounding of vertical profiles of temperature and atmosphere compo- sition. A modification of the ANN method based on minimization of the root-mean-square error of the final determined parameter of the atmosphere is proposed. The significant (by a factor of hundreds) advantage of the proposed technique as applied to the rate of ANN learning and a little advantage in the accuracy of the profile determination are demonstrated by an example of determining the vertical profile of temperature by measurements with an MTVZA device.