Tkachenko R. Feed forward neural networks with non-iteration learning

The thesis is devoted to problems of building, synthesis and application of feed forward artificial neural networks with non-iteration learning.The new conception model of the artificial neural networks have been placed in the methodology base of non-iteration learning. The model "Functional on the tabular functions set" is a functional on the sets of synaptic functions, output functions of neural elements and discrete functions of input sets. The input set is reflected both by network using only projective connections, or including additional lateral ones. This model became the base of a number of new architectures of networks and fast non-iteration algorithms of their learning Created on the basis of provided models and algorithms, program neural networks have found practical applications in the fields of power engineering, and scientific research, as well as in the educational process.