Vol 31, No 6 (2025)

The article proposes the model of a scientific thesaurus, represented as the domain hierarchical graph, which connects the scientific field with scientific terminology. The source of training data for model building is partially structured scientific texts, including subject scientific. We propose the algorithms for calculating the significance of semantic relations between scientific terms. The scientific publication semantic model combines the knowledge stored in the thesaurus with information on term usage statistics in the publication text. The experiments of analyzing scientific publications presented in this paper were conducted using the domain semantic graph "Mathematics", built as a result of automatic processing of the text of the mathematical encyclopedia in five volumes.

Graph neural networks generalize traditional neural networks for graph-structured data and have attracted wide attention due to their impressive capabilities. They are in demand in machine learning, as they can work with heterogeneous information, help to identify relationships between events and data, provide robustness to incomplete, unclear and noisy data, allow to analysis of large amounts of data and structures in the form of knowledge graphs. Training of graph neural networks in hyperbolic space has gained popularity in recent years due to their ability to model graphs with hidden hierarchical data, and because they are more compact and denser. The purpose of applying non-Euclidean geometry theory in modifying graph neural networks is to ensure their stability and better performance compared to Euclidean neural networks. The review of research and development is devoted to the analysis of modern achievements in the field of graph neural networks creation and consideration of accompanying problems. Particular attention is paid to the issues of graph embedding and their representation in non-Euclidean space. This is due to the fact that the performance of models is largely determined by the embedding algorithms and the choice of geometric space for representing graph structures. The prospectivity and efficiency of this approach have been confirmed by works on the creation of hyperbolic convolutional networks of graphs with the property of controlling the curvature of the space in each layer. At the same time, today there is a need for a more detailed consideration of the possibilities of graph neural networks based on the application of methods of hyperbolic geometry because of the insufficient attention paid to these issues in the domestic literature.

A new approach to building secure network channels (tunnels) over the Internet for servicing web services in distributed systems based on the use of the Cryptographic Message Syntax (CMS) standard for secure data representation on the network is considered. Unlike VPN technology, the described approach is strictly focused on supporting only HTTP/SOAP interactions in distributed systems.

It is shown that the described approach makes it possible to include additional functionality in a secure channel. This functionality relates to the demarcation of client access rights to web services and individual service functions, with the routing of information requests in the channel, as well as with the organization of parallel client access to several identical web services at once with the formation of a "total" processing result. The approach implies the use of special gateways that provide encapsulation of HTTP/SOAP-documents into the safe CMS-message structures on the sender side and deencaptulating on the receiver side to make up a "transparent" communication channel for system components. It is assumed that both client programs and web servers are located in the same secure private network (or even on the same network node) with the gateways serving them, and only the interaction between the gateways is carried out through the public network. The implementation of the approach in the Linux and Windows environment and the results of an experimental study are described. In particular, the study showed that when calling service functions with a runtime of 1.0 seconds or higher, the secure channel increases the total query execution time by only a few percent.

Abstract of the article: Grid systems of centralized architecture, with multisite dispatching, characterized by the ability to fulfill a multiprocessor request on several parallel systems simultaneously, are modeled by the resource quadrant. A user request when served by a Grid system dispatcher is modeled by a resource rectangle with horizontal and vertical dimensions respectively equal to the number of time resource units and processors required to complete the request. An illustration of the exponential complexity of dispatching resource rectangles is the placement of successive resource rectangles of equal perimeter from 1*23, 2*22, to 23*1 into an enclosing rectangle of the minimum area, which took more than three days. The exponential complexity of the optimal distribution of resource rectangles determines the practical value of heuristic algorithms of polynomial complexity, which are based on the operations ofdynamic integration of resource rectangles in the resource rectangles environment. To assess the quality of dispatching, a non-Euclidean heuristic measure is used. It takes in consideration the area and shape of the occupied resource area. The quality of dispatching of arrays with requests equal to the non-Euclidean heuristic measure is analyzed. This paper considers the quality of six polynomial level algorithms in terms of height and length (with a disadvantage, with an excess and with a minimum deviation) when dispatching arrays with requests of equal non-Euclidean heuristic measure. The quality of six polynomial algorithms for dispatching arrays with requests of a growing non-Euclidean heuristic measure is investigated. Using five test arrays of resource rectangles with an increasing non-Euclidean heuristic measure, it is shown that Д-level algorithms in terms of length with a minimum deviation have the smallest maximum value of the heuristic measure of 0.7. When servicing arrays with requests of a growing non-Euclidean heuristic measure in Grid systems, it is recommended to use the polynomial Д-level algorithm for length with minimal deviation which is introduced in the paper.

In many scientific fields not only numerical methods, but also symbolic computations tools are used. However, due to the lack of research in the field of performance of symbolic manipulation libraries, it is often hard to determine the optimal software package or library for high-performance scientific research. The purpose of this work is to perform a comparative analysis of various computer algebra libraries for the Julia programming language in terms of computational performance. As part of this goal, a series of benchmarks were developed and carried using the Symbolics.jl, SymEngine.jl, SymPy.jl libraries. Each benchmark was designed to test different common use cases, such as manipulations with large symbolic expressions, differentiation, symbolic expansion, simplification and substitution. Results of each benchmark were analyzed, and the advantages and disadvantages of different libraries were explored. The source code of benchmarks was discussed, and different performance benchmarking features of these libraries were described. In the process of developing benchmarks, various features of computer algebra libraries were discovered, such as effective algorithms for large sum or product generators. The research showed that in most test cases, the SymEngine.jl library demonstrated the highest performance measures. In cases where SymEngine lacks features, SymPy.jl could be used for additional symbolic expression manipulation operations. The results of this paper could help researchers in different fields of science make an informed choice of software library for their tasks.

The paper addresses the problem of educating medical practitioners with digital technologies in the era of healthcare technological transformation. This study describes an interactive simulator which imitates interaction with a digital infrastructure by building a cyber-physical environment. The simulator ensures the development of digital skills through a comparison of processes occurring in biological and computer systems. The results of the simulator testing in describing a thermoregulation process demonstrated the effectiveness of developing in-demand digital skills.