Vol. 21, issue 06, article # 14

pdf Matveev K. A., Pashnev V. V., Pavlov V. E. On applicability of neural network technique to determination of single-scattering particle albedo from clear sky diffuse brightness. // Atmospheric and oceanic optics. 2008. V. 21. No. 06. P. 480-482.
Copy the reference to clipboard

The applicability of neural network techniques to the problem of retrieval of the aerosol particle single-scattering albedo from calculations of the clear sky brightness in a spectral range 0.675 µm is under analysis. A homogeneous neural network consisting of three latent layers, each of 10 neurons, has been considered. A complex of optical parameters covering their actual variations under different atmospheric conditions was used in learning the network. The network was tested in accordance with many samples under learning. A histogram of deviations of the found albedo values from model ones is presented. It follows from the histogram that deviations within 3% are observed in 96.8% of cases.