Statistical and Intellectual Methods for Solution of Inverse Problems
It should be noted that the regularization method requires fitting of the regularization coefficient, which determines, to a great extent, the accuracy of reconstruction of the sought parameters of aerosol particle microstructure. There exist various techniques for automatic fitting of this coefficient, but they significantly decrease the speed of the algorithm because of the need to multiply large matrices many times (see, e.g., References 49, 79 and bibliography therein). In addition, the regularization method quickly loses its stability with an increase of the multiple-scattering background noise in the received signal.
This problem becomes governing in the sensing of optically dense atmospheric hydrometeors, including, in the first turn, low-level stratified clouds, whose monitoring is rather urgent from the viewpoint of flight safety. In connection with the arising difficulties of quantitative interpretation of optical sensing data, the interest in statistical methods of solution of ill-posed problems [76—78] and methods using the artificial intelligence technology [49,79—86] have increased in the recent time.