#3 / 2019

The paper presents data on the fauna of bivalve mollusks of Kabardino-Balkaria. In total only 6 species of the studied group of aquatic organisms have been found in water ecosystems of the Kabardino-Balkarian Republic. Almost all of them are confined to the water bodies of the lowland-foothill zone of the Kabardino-Balkarian Republic.

49-58.pdf

(889 Kb)

Mathematics

We prove that the invariant straight line of the affine-transformed differential system, the right parts of which are n-degree polynomials, passes through the n-th equilibrium of this system, two of which are out-of-node points. At the same time, if the system has two invariant straight lines passing through a certain out-of-node point, then n is odd. Conditions are identified under which the system has no invariant straight line. The main result is the assertion that the number s of invariant straight lines of the system, when certain inequalities are fulfiled, meets the condition s is less than or equal to 2n+4 (n is greater than or equal to 3) , provided that this estimate is accurate at n = 5. All assertions which are being proved are accom-panied by examples.

11-24.pdf

(449 Kb)

We consider the diffusion equation with a fractional derivative with respect to time, when the concentrated heat capacity of a certain quantity is placed on the boundary of the region. As a result, a priori estimates in the differential and difference interpretation have been obtained.

25-31.pdf

(326 Kb)

We continue to study the problem of polynomial curvilinear multiplicative integrals. In this work we reveal structure of subalgebra of multiplicatively integrated matrix functions of the second order. The study is carried out in degrees of polynomial curves.

32-41.pdf

(338 Kb)

In this paper, we consider the boundary value problem for the equation of moisture transfer with a fractional derivative with respect to time. Difference scheme is constructed. As a result, a priori estimates in the differential and difference interpretation have been obtained. An algorithm for the numerical solution of the problem in question is proposed.

42-48.pdf

(332 Kb)

Technical Sciences

The paper presents studies aimed at finding methods for applying quantum neural networks (QNN) for pattern recognition and pattern detection, providing a significant reduction in computation time while maintaining a given accuracy. The features of the modern approach to solving pattern recognition and detection problems based on neural networks and deep learning are considered. The principles of data processing in artificial neural networks are analyzed, and the algorithms used in them are presented. An example of pattern recognition and detection is used to highlight the disadvantages that can be solved with the involvement of quantum neural networks, and to report the possible benefits derived from the transfer of algorithms and processing principles of modern deep neural networks to the quantum computing platform. The feasibilities are analyzed and an assessment is made of the applicability of the use of QNN for problems of pattern recognition and detection and, as a result, of practical, physical implementation of QNN on simulators and later on real quantum devices. A method of teaching QNN for computer simulation of quantum principles of data processing is presented, which will allow studying and analyzing the functioning of the QNN created, as well as optimizing it taking into account the features of quantum computing simulation.

59-68.pdf

(459 Kb)

This article proposes a methodology for calculating the optimal confidence interval and confidence probability for a random sample. For this, a method was used based on maximizing the average increment of information about the sample, which allows taking into account such an indicator as the uncertainty of the initial information. A numerical experiment was conducted in which the relationships between the optimal confidence probability and the confidence interval on the sample size, the distribution law of a random variable, and the type of uncertainty were investigated. Numerical values of the sample size are established, upon reaching which the transition is implemented from the deterministic type of uncertainty to the stochastic type, and then to the fuzzy.

69-74.pdf

(621 Kb)

The paper presents the authors’ method of calculating the parameters of anchor winding and winding coefficients of first harmonic of the generator with excitation from constant magnets of conic construction. The calculation of the generator developed by the authors for work as a part of a hybrid power complex is performed.

75-84..pdf

(743 Kb)

The article describes the use of deep Siamese convolutional neural networks for solving problems of detecting and identifying people with a video stream in conditions of limited data in order to reduce the complexity of models and the number of errors at the learning stage. Biometric data of the person is converted into a compact size – embedding. To solve detection problems, a method is applied based on the calibration of evaluation vectors; in this area, the use of methods for normalization and L2 regulation with loss of triplets is shown. For input data, a batch normalization is performed; L2-regularization minimizes large weights while maintaining parameters that are not subject to retraining; in turn, the loss of triplets is used to minimize the Euclidean distances between embeddings. The use of these approaches allows achieving high accuracy rates with a rather non-standard perspectives and uneven illumi-nation of facial images.

85-90.pdf

(533 Kb)

The article substantiates the relevance of the application of linear optimization methods to the problems of manufacturing two-three-and multicomponent alloys. The formalized problem of optimization of cost of alloy production taking into account its chemical composition and cost of components is presented. The algorithm of graphical visualization of solution search is described. The results of testing the developed Configurator parameters of the objective function are given.

91-95.pdf

(871 Kb)

Solving the problems of applied photogrammetry with the use of mobile technologies involves the availability of a photo-recording device and a computer system which performs processing and analysis of images that interact with each other through network technologies. A key factor in the construction of systems based on this approach is the inverse relationship between the quality of the photo-recording device and the computational complexity of image processing and analysis algorithms, which limits the mobility, flexibility, ease of operation and minimum cost of the measuring system. The existing software components of photogrammetric complexes are a bundle of preprocessing, pattern recognition and solution of counting problems in their different variations. Despite the presence of such a basic structure, software complexes are built heuristically, and the set and sequence of interaction of means are static. The article proposes algorithmic support of mobile-cloud system, which will minimize hardware requirements through the use of cloud services and technologies.

96-102.pdf

(615 Kb)

The article deals with the availability of electrical industrial oil refining complexes and possible promising directions for the creation of a new type of energy-saving installation for oil refining.

103-108.pdf

(667 Kb)

Automated control systems are an integral part of the IT infrastructure of any enterprise. Currently, developers are using data mining systems. The work presents modern automated enterprise management systems, directions of development of these systems, as well as feasibility of their joint application.

109-112.pdf

(307 Kb)