Bokotsei I. Digital and pulsed neuroelements for patternal neurocomputers

The object of research is the neuroelements functioning processes. Neuroelement is a computer component, which performs the functions of information processing like biological neuron. The study dedicated to improve the neuroelement operational performance by improving their structural and functional organization and implement them on the modern optoelectronic elementary basis. To achieve this goal the mathematical models and structures of neuroelements were improved and the hardware implementation of these neuroelements and compact optoelectronic pulsed neural network were developed. The scientific novelty is that logic-temporal model of the neuron and its functional structure were proposed first. The model used logic-temporal information representation and algebraic summation of time intervals with arbitrary start and end points, that reducing hardware costs and extending the neuroelement functionality. The mathematical model and the structural organization of multibit digital neuroelement were improved. The spatio-temporal summation by difference slices was applied to accelerate the addition operation, that allowed to increase the average neuroelement performance. Structural organization of the pulse neuroelement was Improved by introducing a separate photodetectors for excitatory and inhibitory signals and threshold device with high output current, which allows to build on their basis the optoelectronic pulsed neural networks with improved characteristics. Mathematical model of pulsed neuroelement based on bispin-device was improved, taking into account the internal parameters of semiconductor bispin-structure, that allowed to improve its accuracy, reliability and expand its use to provide the required parameters neuroelements for development and production of semiconductor chips. The practical value is the creation of structural and functional hardware circuit diagrams of the proposed neuroelements and receiving the results of computer modeling and experimental researches of suggested neuroelements. The degree of implementation - the results of the thesis are inculcated in the Zhmerynkavodokanal for underground pipe damage location determining (Zhmerynka, Ukraine), and theoretical principles are used in the educational process at the department of Computer Science. Field (area) of use - neurocomputer processing systems and recognition of signals of different nature, the construction of functional units for pattern computers.