|
The Bulletin of the Adyghe State University, the series “Natural-Mathematical and Technical Sciences” |
 Šóńńźčé
|
| Home page Subjects Chief Editor Editors Instructions to Authors Rules of the reviewing Subscription Contacts Publishing house | ||
|
#4 / 2024Technical SciencesAleksey N. Samoylov1, Andrey I. Kostyuk2, Eduard V. Melnik3, Aleksandr V. Kozlovsky4, Ali M. Garyagdyev5Identification of surroundings objects for drone motion problems in the perimeter monitoring system of agricultural land
The paper considers architectural and algorithmic principles of creating a system of identification of environmental objects for the tasks of mobile drone movement in the system of monitoring the perimeter of agricultural land. The construction of the identification system involves hybridisation of object detection algorithms (YOLO, Faster R-CNN) and machine learning with the use of support vector method (SVM) for classification of detected objects. The practical significance of the proposed solution lies in the autonomous control of drone movements in case of satellite fixation violation and lack of communication with the operator's console. The flexibility of the system is ensured by a multi-component modular layout, which additionally makes it possible to adapt the system to a specific set of tasks.
Using the “PHYPHOX” application to determine the moment of inertia of a cylinder
The paper discusses the possibilities of using the application of the application “Phyphox” for smartphones to determine the moment of inertia of a hollow cylinder of different diameter. The experimental dependence of the moment of inertia of the body on the radii of cylinders is obtained. The moment of inertia of the cylinder is determined by the method of rolling down an inclined plane. An example of using the “Phyphox” application option for obtaining experimental data of velocity and angular velocity of the body at an arbitrary moment of time is given. As a result of the experiment, the values of the moment of inertia at the cylinder mass m=0,225 kg and radius R1=0,0425 m J1=4,06·10-4 kg·m2 are obtained. The measurement error in the first experiment is 3,5 %. At mass m=0,225 kg and radius R2=0,05 m J2=5,51·10-4 kg·m2. The measurement error in the second experiment was 2,2 %. At mass m=0,225 kg and radius R3=0,05 m J3=8,79·10-4 kg·m2. The measurement error is 7,9 %.
Development and research of error backpropagation algorithm for quantum network
The paper considers the algorithm of error back propagation adapted for quantum neural net-works. The main attention is paid to the development and modification of the traditional method of error back propagation for its application in conditions of quantum nature of data. Theoretical aspects related to quantum states, superposition and entanglement, as well as their influence on the efficiency and correctness of training of quantum networks are analyzed. The relevance of the mathematical algorithm, including the use of density matrices and quantum operators to describe and optimize the network parameters, is substantiated. Experimental modeling of the proposed approach on quantum data sets is carried out, which showed high accuracy and stability of the proposed method. The limitations of the algorithm related to noise in quantum sys-tems and computational costs with increasing number of qubits are discussed. The results demonstrate the potential of the developed algorithm for optimization and training of quantum models, which opens new perspectives for further development of quantum machine learning and creation of hybrid classical-quantum algorithms.
Systematic analysis of basic methods of neural network protection of information from DDoS attacks
The main problems associated with the threat of DDoS attacks are reviewed and various approaches to solving them using neural networks are proposed. The study covers various aspects of information security, including detection and classification of DDoS attacks, traffic analysis, and prediction and prevention of possible attacks. The authors analyze both classical methods and innovative approaches based on deep learning and artificial intelligence. The results of the study allow to identify the most effective methods of defense against DDoS attacks using neural networks and learn the prospects for their improvement. The presented analysis is important for developers of information protection systems against DDoS attacks and cybersecurity special-ists.
Automated system-cognitive analysis of the company's information security status
The paper is devoted to the creation of an adaptive tool for solving the problems of predicting and minimising the consequences of vulnerabilities in the security settings of computers running under MS Windows operating systems. The choice of automated system-cognitive analysis (ASK-analysis), which allows solving the above problems, is justified, its ability to process fragmented and noisy data of different nature is emphasised. Intelligent system “Eidos” provides processing and analysis of large amounts of data related to the configuration of system security, allows to predict potential risks and provides support for decision-making on their minimisation. The pa-per includes a description of the stages of ASK-analysis from cognitive structuring of the subject area to model synthesis and decision support. The developed methodology is well suited for im-proving information security of small and medium-sized business computer systems.
Problems of training and using of artificial intelligence in medicine
The article systematizes in tabular form the main problems of the use and training of artificial intelligence systems in healthcare.
Structure of GSM mobile networks
In the paper the structure of GSM mobile networks is considered. The network architecture, including the radio subsystem, base station subsystem and switching subsystem are analyzed. The different cell types used in GSM networks and their impact on coverage and capacity are discussed. A detailed overview of the protocols used in GSM networks is presented, including radio channel access protocols, mobility management protocols, and signaling protocols. Security mechanisms implemented in GSM networks, such as encryption and authentication, are analyzed. The advantages and disadvantages of GSM networks are investigated, including their wide coverage, reliability and relatively low cost. The limitations of GSM networks such as lim-ited bandwidth and vulnerability to certain types of attacks are discussed. The paper is intended for engineers and students interested in mobile cellular networks. It provides an understanding of the structure and operation of GSM networks, which is an essential element in the design and optimization of future mobile communication systems.
|
© Adyghea State University, 2005-2023