Neural networks in the problems of adaptive optics and recognition of optical images obtained in the presence of phase fluctuations

The authors describe several architectures for adaptive optics which use neural networks and discuss their advantages and drawbacks. The introduction of special processing of light fields and training neural networks for the joint recognition of intensity (amplitude) holograms and phase components of the received waves makes it possible to increase the efficiency of recognition of objects observed in the presence of phase fluctuations rapidly. Application of little-modified standard algorithms which are usually used for training the neural networks make it to use neural nets for efficient adjustment of the wavefront in architectures of adaptive optics and to improve the quality of images distorted by phase fluctuations