Методи та засоби паралельних перетворень векторних масивів даних

Abstract

The thesis is devoted to performance capabilities improvement and expansion of vector processing methods of the large data array under simultaneous processing with neurostructures and neurocalculations orientation. In the work the method of parallel addition of the numerical data using differential shearing principle is improved, thus the functionalities of the given method raise. The requirements are formulated and the practical realization examples of the given process are quoted. The time law of elements multisummation of vector data array are investigated and proved. The following simultaneous processing of such array methods are offered: parallel algebraic multisummation as a result of improvement the known parallel addition of the numerical data; parallel threshold algebraic multisummation as a result of which research the formal neuron model is improved. The closure method elements of vector data array into circuit is investigated during sorting by the pair swap method; the mathematical model of the given method is offered and the examples of practical implementation are quoted.