Described is a possible approach to solving the problem of obtaining a forecast of output parameters of cellular structures with those combinations of values of parameters of the object which are in previously given ranges based on studies already carried out, based on use of previously obtained results of variant calculations. Provides portions of a suite of programs that use MATHCAD to implement this approach. Comparisons of methods for eliminating heteroskedasticity with different selection of weighting factors based on the use of the weighted least squares method were made.
The article presents an algorithm for automatic extraction of metadata from executable software files at the disassembly stage. Methods for reading binary content, converting byte sequences into string representation, and filtering the obtained data using PowerShell and strings.exe utility are considered. An experiment was conducted confirming the possibility of extracting structural and textual elements of executable modules without their complete decompilation. The aim of this work is to improve the efficiency of analyzing binary applications by highlighting internal structural data at an early research stage. Methods used include static analysis, reverse engineering, text analytics, PowerShell scripting, and command-line tools. Results achieved include a working algorithm capable of extracting basic metadata characterizing the structure and functional dependencies of executable code. The scientific novelty of the research consists in the following points: A new algorithm has been developed for extracting metadata from executable files without decompiling or loading them into a disassembler; A method of binary analysis with sequential transformation of byte arrays into text form with data filtration has been proposed; Integration of PowerShell and strings.exe utilities into a unified analysis scheme ensuring compatibility and automation has been implemented; Classification of extracted data into textual, system, and structural elements has been defined; Textual analytics technique has been applied for the first time in the context of binary analysis, thus expanding reverse engineering methods.
The neural network for diagnosis is trained on international data from patients with neurological diseases (ND). The neural network (NN) is dynamically updated with data on the condition of a particular patient to support therapy. An IoT device (smartphone) collects patient data (voice, movement), initiating their preprocessing and feature extraction. This data is transmitted via the local Flask framework to a server using Open Semantic Technology for Intelligent Systems (OSTIS), a platform that processes and interprets patient information. An NN-based forecasting agent on the server, using a trained neural network, performs the patient's diagnosis, which is transmitted to the doctor. Treatment of ND requires continuous, adaptive and individual therapy. The NN-based electronic therapy unit, which includes modules for patient and doctor, decision-making and the patient's digital twin, expands the diagnostic system to support therapeutic decisionmaking. The block searches for the patient's previous condition; predictive modeling: course of treatment, drug therapy; explanation for the doctor. The block structure includes: a recurrent neural network based on GRU, a component of the semantic memory of the OSTIS system, context-sensitive rules in the form of SC-graphical forms.
The article is devoted to the study of new architectures of artificial neural networks aimed at improving classical transformers by reducing computational costs and increasing efficiency. The main stages of the evolution of transformers are considered, starting with the introduction of the mechanism of self-attention and ending with modern sparse models. Particular attention is paid to optimization techniques such as parametrically efficient fine-tuning (PEFT), adaptive layers (adapters), as well as the use of external knowledge repositories and memory vectors to expand the long-term memory of models. The article concludes with a discussion of current achievements, limitations, and promising areas for future research in this area.
The article examines modern approaches to forming an optimal investment portfolio in the cryptocurrency market, including classical methods of technical analysis. Trends in the development of the global digital financial assets market are studied. Various interpretations of the concept of investment strategy are considered. Particular emphasis is placed on the Markowitz model. Its basic version and modification for crypto-futures and crypto-options are reviewed. A computer implementation of the model was developed, numerical experiments were conducted, followed by a substantive interpretation of the obtained results. An algorithm for solving the nonlinear optimization problem using pre-trained neural networks and the Pyomo framework was developed. Additionally, alternative approaches to cryptocurrency portfolio optimization are considered.
This article examines the general principles of constructing a decision support system for controlling an unmanned aerial vehicle (UAV) from a ground control station based on measurements and assessment of the technical condition of the UAV's onboard systems. An analysis of current methods for solving such problems is provided, and the use of algorithmic solutions based on stochastic mathematical models of the processes of change in variables characterizing the UAV's technical condition is substantiated. General differential equations describing the evolution of the system under consideration are presented, derived from the theory of systems with random structure changes. A block diagram characterizing the general principles of solving the problem of probabilistic assessment of the system's technical condition is presented, based on which a discrete algorithm for solving the problem of optimal filtering and identification of the operating mode with monitoring process control is derived. An example of solving the problem for simple mathematical models of the process of change in a technical parameter and its measuring device is given, clearly demonstrating the operability of the proposed algorithm.
There is a need for legislation, guidelines, and organizational structures in the management of Information Systems (IS) in South Africa’s public sector to ensure that technological solutions support strategic direction. This study synthesizes feedback received from 55 GITOs (Government IT Officers) to understand tools that shape IS operations and corporate governance. The findings provide a complex picture of governance and can be regarded as follows: There are acts such as the Financial Management Acts and ICT laws, which set standards dealing in accountability and security other frameworks include the DPSA’s: Corporate Governance of ICT Policy Framework (CGICTPF), the King IV, COBIT, and ITIL. These principles are given effect through internal governance instruments such as the ICT steering committees and the audit committees. The findings are presented using bar graphs, which indicate that more than half the identified GITOs consider essential tools such as the CGICTPF, PFMA, and audit committees. These findings of the study are connected to the previous literature with considerations of comparability with international standards and areas of concern, including resource limitation as well as the trade-off between conformity and creativity. Finally, it outlines recommendations for improving the IS governance in the context of the South African public sector, including the need for constant improvement of frameworks and the improvement of IS leadership competencies for public sector IT professionals.
We benchmark practical ways to accelerate Stable Video Diffusion (SVD) inference for image-to-video. All methods use a fixed setup (1024×576 input, 25 frames) with an FP16 baseline at 25 denoising steps. We compare eight techniques: UNet INT8 weight-only quantization, torch.compile+TF32, step reduction, distilled weights (AnimateLCM), semi-structured 2:4 sparsity, LCM mode/scheduler, and keyframe generation with RIFE interpolation as post-processing. We measure latency and peak VRAM, and track quality/motion via CLIP similarity, tSSIM, and tLPIPS. Keyframes+RIFE achieves the highest speedup while preserving strong conditioning to the input. LCM provides a balanced ~2× speedup, whereas aggressive step cuts (and untuned 2:4) can degrade motion.
In this article, the necessity of applying a systematic approach to ensuring safety in the grain drying process at grain drying complexes and ensuring grain quality, as well as the efficiency of the drying process, is justified. An analytical mathematical model of the drying process has been developed in the form of a system of differential equations that describes the dynamics of temperature change in the grain drying process.
This article analyzes the potential of the existing Bluetooth detector network for constructing traffic models and analyzing traffic flow redistribution. It also attempts to validate readings from stationary radar detectors by comparing them with reference data obtained through video analysis. This article monitors traffic detector readings and assesses the compliance of traffic detector data received by the automated traffic management system and the VCC platform with actual data. The technology links traffic detectors to control actions based on the developed traffic management technology. Suggestions are made regarding the need for calibration and adjustment of traffic detector positions. A roadmap is proposed for addressing the identified deficiencies and improving the quality of data, which is the foundation for safe and effective traffic management within the ITS framework.
The subject of research is the analysis of the use of calculation automation tools for the use of factorial, regression, and correlation analyses. The purpose of the article is to present the well‒known and author's experience of using the Python language for applying factorial, regression and correlation analyses and the experience of the authors in their application. The main aspects of the theoretical foundations of these methods of analysis are considered. The classification of software tools for automation of statistical calculations is given. The advantages and limitations of programming languages (Python, R), commercial platforms (SPSS, SAS, Stata), BI platforms (Tableau, Power BI, Zoho Analytics), open platforms (KNIME, RapidMiner), specialized statistical packages (JASP, Jamovi) are considered. The methodology of the application of automated statistical analysis is detailed, including the preparation of initial data, the choice of method, tincture of parameters, performing calculations and obtaining primary results. The author's methodology for automating complex analysis for evaluating the use of neural networks, the Internet of Things and blockchain in the university's educational process is presented.
The article presents the results of the development and testing of a prototype automated system for remote screening of diabetic retinopathy (DR), featuring preprocessing and classification functionalities for digital images acquired by fundus cameras from various manufacturers. Experimental testing of the prototype was performed on several thousand real-world images from Belarusian medical institutions and open-access sources. Objective assessment of retinal image quality is computed through analysis of the central fragment based on the Weibull distribution scale parameter applied to local quality estimates of the green channel of the image. The prototype was developed on the Streamlit platform and employs the YOLOv8 and YOLOv11 neural network architectures for classifying five stages of DR, achieving classification accuracy of 84.0 % and 82.7 %, respectively. The combination of low-quality image filtering and neural network-based classification enhances the reliability of screening diagnostics and can be deployed in telemedicine systems with limited computational resources.
In the paper we present a comparative study of binarization methods for color images of static signatures made with ballpoint pens of different types and colors. Signature image binarization is the first step before calculating its features and verification. Because of the uneven flow of ink from ballpoint pens, images of signatures made with such pens present particular challenges. A comparison of digital signature image preprocessing methods aimed at preserving the shape of the signature lines in the binary representations is conducted. A comparative analysis of binarization methods for color signature images is performed based on four methods from different classes: global thresholding (Otsu, Kapura), locally adaptive thresholding (Sauvola), and a method of direct indexing the RGB color space into two classes: white and black pixels. For the first time, empirical objective criteria for the quality of a binary signature representation in the absence of a reference are proposed, based on the analysis of connected components and the skeleton of the binary signature representation. Experiments were performed on images from the publicly available CEDAR database and a database of signatures collected during the research. It has been shown that Kapur's method provides the best preservation of signature form in its binary representation, outperforming other methods, including the popular Otsu method. We propose a four-step procedure for generating a binary signature representation. This procedure consists of scanning a color signature at 300 or 600 DPI in the RGB model, converting the color image to grayscale using principal component analysis (PCA), binarization by the Kapoor method, and post-processing the binary image. This method is intended for developing static signature verification systems.
While use algorithms of pre-processing raster image involves improving image characteristics to use qualitative transformation effect result on following processing stages. Assessing the effect of an impact on a raster image is not a trivial task, where the choice of further processing steps and algorithmic tools depends on the quality of its implementation. The algorithm for balanced "extension" of boundary pixels of a static raster image is as example. Algorithm developed to improve the results of object edge detection algorithms based on the two-dimensional convolution operation, where the improvement consists of reducing the number of sharply contrasting pixels at the boundaries of raster images relative to adjacent ones. Existing numerical raster images estimates of PSNR and SSIM which obtained after applying the balanced algorithm of redefining border pixels have shown their low efficiency due to the small number of pixels affected by the transformations. The tonal deviation of raster image boundary pixels is alternative assessment criterion. The object of study in the article is the pixel matrix of three-channel color static raster image. The subject of the study is the evaluation of the tonal deviation of the boundary pixels of a static raster image obtained by used balanced algorithm for redefining border pixels. The main objective is to calculate the tonal deviation estimates of boundary static raster images after applying balanced algorithm of "redefining" border pixels, analysis and ranking by alternative methods of "extension" of boundary sizes according to the criterion of tonal deviation. The article provides a definition of the concept of tonal deviation and justification for the use of tonal deviation assessment. An example of calculating the tonal deviation of pixels after applying the balanced algorithm of "redefining" border pixels and the copy method for 1 thousand color three-channel raster images from the SpaceNet Satellite Images dataset from ImageNet is presented. It was carried out short analysis of the obtained assessment results and recommendations for choosing the best methods of "redefining" border pixels.
This scientific article presents a comprehensive theoretical and applied study dedicated to the optimization of atmospheric air monitoring systems in modern cities with populations over 100,000. The relevance of this study stems from the need to transition from traditional passive static monitoring methods to dynamic adaptive models of environmental quality management, a key element of the "smart" and "symbiotic" city concept for minimizing environmental risks. Four major industrial and administrative centers of the Republic of Belarus – Grodno, Pinsk, Lida, and Soligorsk – were selected as the research sites. The scientific novelty of this study lies in the development of an integrated approach combining a mathematical solution to the optimal sensor placement problem (SLP) with a multivariate graphical analysis of anthropogenic impact. A detailed methodology is provided for conducting 20,000 iteration cycles for each transport scheme, which made it possible to identify critical nodes in urban framework networks where the concentration of sensors provides minimal error in calculating transport and environmental parameters. Particular attention is paid to the interdisciplinary consideration of meteorological factors, in particular, the use of a weighting factor for wind rose indicators when prioritizing gas analyzer installation locations. The article systematizes data on traffic load, the hierarchy of street and road networks, and the density of social facilities in the studied cities. Implementation of the proposed model (based on 32 automated posts) allows for the generation of symbolic solutions for assessing the environmental situation in unmonitored network sections. The study's results form the basis for the development of predictive emission management systems capable of adjusting traffic management in real time to prevent exceeding maximum permissible pollutant concentrations. The conclusion outlines promising areas for integrating the resulting models with machine learning and artificial intelligence algorithms within the framework of longterm sustainable urban development programs.
The purpose of the research is to develop additional capabilities of the cybernetics methodology for conducting numerical assessments and analyzing the state of management, control and safety of technical systems. The purpose of the study is to formulate and present elements of a methodology for analyzing quantitative patterns for ergatic systems, characteristics of projects of technical systems, equipment, devices, technologies for various fields of industry, nuclear energy, which can be used to describe their conditions, compare projected, planned or applied objects of study based on data from their technical specifications, passports and procedures., without using long-term experimental or operational data. The quantitative pattern analysis approach proposed on the basis of the ideal technical system model makes it possible to evaluate and analyze the advantages of innovative technologies, to make a comparative assessment of the state of management and safety of technical complexes, to form justifications for the preparation of optimal measures for programs and plans for the implementation of experimental and commercial technical projects, their control, modifications and upgrades, to develop recommendations for a comprehensive analysis in during the examinations. The features of the application of the equation of state hypothesis for the analysis of real technical systems are presented. The use of auxiliary approaches to evaluate the characteristics of technical systems allows for the development of reasoned management decisions based on a comprehensive analysis of a larger number of properties and quantities under consideration.
The exponential growth in the size of Large Language Models (LLMs) creates significant barriers to their local deployment, primarily due to Video RAM (VRAM) shortages on single devices. The aim of this work is to identify and substantiate the most effective parallelism strategy for LLM inference on consumer Graphics Processing Unit (GPU) clusters connected via a slow PCIe bus. Research methods included a series of experiments comparing a monolithic architecture (NVIDIA RTX A6000) and a distributed system (2x NVIDIA RTX 3090) using the vLLM framework. The impact of Tensor Parallelism (TP) and Pipeline Parallelism (PP) on key metrics – throughput, latency (TTFT, TPOT), and power consumption stability – was analyzed while running the DeepSeek-R1-Distill-Llama-14B model. The results unequivocally indicate the unsuitability of Tensor Parallelism for systems without NVLink due to critical synchronization delays. It is proven that Pipeline Parallelism is the only viable strategy for PCIe clusters, ensuring high throughput despite the presence of idle periods («bubbles») and a less stable power consumption profile compared to the monolithic solution. In conclusion, recommendations for using multi-GPU configurations are formulated: they represent the optimal economic choice for memory-critical tasks, such as Retrieval-Augmented Generation (RAG), allowing VRAM scaling at a significantly lower cost than professional analogs.
This review article presents a generalized, integrated classification of scoring functions for molecular docking of protein interaction models. It describes various types of scoring functions and reports relevant issues in biopharmaceuticals, medical cybernetics, computational biology and biophysics, applied mathematics. This article may be of interest to a wide range of readers, including scientists from various fields, software engineers, developing automated bioinformation processing systems, lecturers on relevant topics, graduate and postgraduate students in bioinformatics, systems analysis, medical software systems research
The article examines an approach to the design and analysis of the organizational model of production management at an industrial enterprise based on cognitive and simulation modeling. The relevance of the study is due to the need to improve industrial enterprise management systems under conditions of high external environmental variability, increasing complexity of production and managerial relationships, and the insufficiency of static optimization models for describing the dynamics of organizational processes. The aim of the study is to formalize the organizational model of production management at a metallurgical enterprise using a causal loop diagram and to implement it subsequently in a simulation modeling environment. The research employs methods of systems analysis, economic-mathematical modeling, cognitive modeling, and system dynamics modeling. The organizational structure of the enterprise is presented as a set of production and functional subsystems interconnected through vertical and horizontal managerial interactions, while the key parameters of the model include span of control, management personnel maintenance costs, production capacity, sales volume, and net profit. The constructed cognitive map made it possible to identify the main cause-and-effect relationships among the organizational, production, and financial parameters of the enterprise, while the simulation model enabled an assessment of the impact of managerial decisions on the dynamics of costs and financial performance. The results obtained show that the use of cognitive maps in combination with simulation modeling expands the possibilities for analyzing the stability of the organizational structure and for substantiating decisions aimed at its improvement. The practical significance of the study lies in the possibility of applying the proposed approach to the design and adjustment of management systems in metallurgical and other industrial enterprises.
This article examines the general principles for assessing the effectiveness of unmanned aerial complexes (UAC), including groups of unmanned aerial vehicles (UAVs). It analyzes current methods for assessing effectiveness in performing various UAC tasks. Key performance indicators are substantiated as probabilistic characteristics. General probabilistic performance indicators and specific indicators characterizing individual stages of UAC use are formulated: the probability of object detection by a UAV group and the probability of timely resolution of the task of interfering with the object (providing assistance to victims), taking into account time delay characteristics. Particular attention is paid to solving problems of searching for objects in a given area of the Earth's surface due to the relevance and prevalence of such problems. Mathematical relationships are presented for determining the probability of finding an object in a given area, determining the probability of detecting an object by a given deadline, and calculating the mathematical expectation of the number of objects detected by a given deadline.
