The hypothesis about the agreement of theoretical and empirical Pareto distributions is tested in relation to cumulative curves - diagrams for instructions and microinstructions of an educational computer. The Kolmogorov criterion is used as a statistical criterion for agreement. The values of the parameters are obtained for the Pareto distribution functions that describe the probabilistic properties of random variables, which are the ordinal numbers of commands and microcommands implementing them. Constructed graphs of theoretical and practical distribution functions make it possible to exclude rarely used commands from the computer command system, which helps to simplify the architecture of computer processors. Certain theoretical principles and the obtained practical results are a further development of the statistical method of improving quality, i.e. Pareto analysis, in relation to quantitative assessment of metrics of machine commands and microcommands of a computer.

The correctness of using the multi-seasonal season-trend decomposition method based on locally weighted scattergram smoothing for the problems of forecasting multi-seasonal load processes in elastic systems is assessed. A comparative analysis of the performance and accuracy of the above method and the seasonal integrated autoregressive moving average (SARIMA) model is performed. Results of experiments are presented that confirm the difficulty of constructing the SARIMA model based on data with a high degree of discretization and period values exceeding classical seasonality, such as 7, 12, 52. When creating the SARIMA model, time restrictions are imposed on the selection of parameters due to high memory consumption, which lead to a decrease in forecast accuracy and limited the ability to build a model based on higher seasonality indicators. The multi-seasonal season-trend decomposition method demonstrates an advantage over the SARIMA model in terms of forecast execution time and memory consumption, however, with a small set of initial data, the SARIMA model shows higher accuracy.

Results of a study on creation of a methodology for assessing the operation system of complex object based on a fuzzy-possibility approach using explicit and implicit professional expert knowledge are presented. The factor space contains seven fuzzy linguistic variables for constructing a fuzzy-possibility mathematical model for assessing the state of objects of ground-based space infrastructure, in which the space tracking and surveillance system is considered as a complex object, and its dependence on the quality of functioning of the corresponding system is also studied. The effectiveness of the created methodology was assessed based on calculations using the constructed fuzzy-possibility model.

The deployment of Convolutional Neural Networks (CNNs) models on embedded systems faces multiple problems regarding computation power, power consumption and memory footprint. To solve these problems, a promising type of neural networks that uses 1-bit activations and weights emerged in 2016 called Binary Neural Networks (BNNs). BNN consumes less energy and computation power mainly because it replaces the complex heavy convolution operation with simple bitwise operations. However, the quantization from 32-float point to 1-bit leads to accuracy loss and poor performance, especially on large datasets. This article presents a review of the key optimization techniques which influenced the performance of BNNs and led to higher representation capacity of BNN models, as well as an overview of the application methods of BNNs in object detection tasks and compares the performance with the real value CNN.

The problem of computer simulation of an optical system with light scattering elements is considered. Correctly modeling devices with such elements requires lengthy light calculations using stochastic ray tracers. An approach to efficient modeling of such installations is presented using the example of a patented device designed to measure the bidirectional scattering function, which is used to describe light scattering properties. A realistic computer model of such a device is proposed, which makes it possible to calculate tolerances for deviations in the positioning of the most critical device blocks, and an assessment of the accuracy of its measurements is carried out. Results of measurement simulations of bidirectional functions are presented in the form of graphs and synthesized images.

An arithmetic calculator that performs addition and subtraction operations with numbers in a fixed-point format in direct code and is built on elements of neural logic, is presented. Arithmetic operations of summation or subtraction are performed by analyzing the operation code and signed digits of numbers. If the sum of the sign bits of binary numbers and the operation code is equal to zero, then the summation operation is performed, otherwise - subtraction. When summing, the transfer from low to high digits is determined; when subtracting, a loan from high to low digits is calculated. The proposed device has increased computing speed, and the use of neural-like elements allows reducing hardware complexity.

Research has been carried out to determine the effective area of the photosensitive surface of silicon photomultipliers (SiPM) at the operating supply voltage, as well as to establish the dependence of this characteristic on the applied supply voltage. Silicon photomultipliers KOF5-1035 (Belarus), Ketek RM 3325 and ON Semi FC 30035 (Germany) are selected for the study. Application of developed experimental setup makes it possible to reveal that an increase in the supply voltage leads to an increase in the area of the effective photosensitive surface of the SiPM. It is found that the maximum value of sensitivity at a given operating voltage is observed in the central part of the photosensitive surface and decreases symmetrically as the spot of the optical probe approaches the surface edge. The results obtained can be used to create devices for detecting optical radiation in the visible range based on silicon photomultipliers.

An approach to the automated acquisition of non-contrast computed tomography (CT) images containing abdominal aortic markings derived from contrast-enhanced phase scanning data is presented. An algorithm for suppressing contrast enhancement in the area of the abdominal aorta on a CT image is developed. The scientific novelty of the approach lies in the conversion of marked contrast images into non-contrast images using a developed mathematical model that allows for isolation and suppression of the component of X-ray absorption of the contrast agent. The algorithm was tested on an open data set consisting of 4 CT studies of the abdominal aorta, the balance of “aneurysm: normal” classes was 1:1. The results demonstrate the comparability of the X-ray density values in the study area with literature data, as well as the similarity of this area with the surrounding muscle tissue. Expert classification of a mixed sample containing real and generated images demonstrates the realism of the latter (accuracy of detection of artificial images - 35%, Fleiss kappa - 0.12). The resulting images are intended for training and testing artificial intelligence algorithms in the field of opportunistic screening of aortic aneurysm.

Deformation of a 3 mm thick rubber membrane under pressure created by working fluid is considered. The membrane is installed inside a cylinder and supported by a piston. When the piston is lowered, the membrane moves with it and the pressure of the working fluid deforms the membrane, stretching the rubber along the edges of the piston. A solution to the problem of determining the breaking stresses is presented, with special attention being paid to the area close to the membrane fastening, where the main deflection of the rubber and its inversion to the other side occur. Calculation and analysis of the results obtained of the area in the corners of the membrane along the edges of the piston, where the rubber is stretched under pressure, are performed. The problem is solved as a static one in an axisymmetric formulation without taking into account friction using the ABAQUS program. The most dangerous option is considered with zero friction between rubber and metal, and rubber is considered as a nonlinearly elastic, weakly compressible material. The results of numerical calculations on the strength criterion based on standard rubber indicators: tensile strength and elongation at break are analyzed. A conclusion is presented about the performance of the membrane according to the strength criterion.