The neural network approach in construction of adaptive grids for irregular flat areas

The paper introduces the modified algorithm of construction of adaptive grids on complex irregular flat areas which can have branching narrow cracks, multiple connection, cuts of different sizes and etc. Program realization of this method is carried out and the program interface allows an interactive variation of parameters of training neural network. The improved adaptation of networks on irregular areas will be used in problems of numerical modeling of physical processes, processing of images, visualization of data, graphic appendices and etc.