AN APPROACH TO ACCELERATE THE TRAINING OF CONVOLUTIONAL NEURAL NETWORKS BY TUNING THE HYPERPARAMETERS OF LEARNING

Over the last decade, a set of machine learning algorithms called deep learning has led to significant improvements in computer vision, natural language recognition and processing. This has led to the widespread use of a variety of commercial, learning-based products in various fields of human activity. Despite this success, the use of deep neural networks remains a black box. Today, the process of setting hyperparameters and designing a network architecture requires experience and a lot of trial and error and is based more on chance than on a scientific approach. At the same time, the task of simplifying deep learning is extremely urgent. To date, no simple ways have been invented to establish the optimal values of learning hyperparameters, namely learning speed, sample size, data set, learning pulse, and weight loss. Grid search and random search of hyperparameter space are extremely resource intensive. The choice of hyperparameters is critical for the training time and the final result. In addition, experts often choose one of the standard architectures (for example, ResNets and ready-made sets of hyperparameters. However, such kits are usually suboptimal for specific practical tasks. The presented work offers an approach to finding the optimal set of hyperparameters of learning ZNM. An integrated approach to all hyperparameters is valuable because there is an interdependence between them. The aim of the work is to develop an approach for setting a set of hyperparameters, which will reduce the time spent during the design of ZNM and ensure the efficiency of its work. In recent decades, the introduction of deep learning methods, in particular convolutional neural networks (CNNs), has led to impressive success in image and video processing. However, the training of CNN has been commonly mostly based on the employment of quasi-optimal hyperparameters. Such an approach usually requires huge computational and time costs to train the network and does not guarantee a satisfactory result. However, hyperparameters play a crucial role in the effectiveness of CNN, as diverse hyperparameters lead to models with significantly different characteristics. Poorly selected hyperparameters generally lead to low model performance. The issue of choosing optimal hyperparameters for CNN has not been resolved yet. The presented work proposes several practical approaches to setting hyperparameters, which allows reducing training time and increasing the accuracy of the model. The article considers the function of training validation loss during underfitting and overfitting. There are guidelines in the end to reach the optimization point. The paper also considers the regulation of learning rate and momentum to accelerate network training. All experiments are based on the widespread CIFAR-10 and CIFAR-100 datasets.

WEB TECHNOLOGY OF USING VIDEO CAPTURE PROGRAMS IN THE PREPARATION OF EDUCATIONAL MATERIALS FOR DISTANCE LEARNING

Promising areas for improving the education system, and with the advent of coronavirus online mode has become critical in education, in this regard are at home. Schools have started using Zoom software, and the role of the modular learning environment has grown significantly. At the same time, many problems were identified, such as the lack of Internet access for students from rural areas or disadvantaged families, which has become an obstacle to learning and access to educational materials. In response, the United Nations Educational, Scientific and Cultural Organization has made a number of recommendations for distance learning. Modern distance education is an extensive system of knowledge transfer at a distance using various tools and technologies, the result of which is the receipt by users of the necessary information for its successful implementation in practice. There are a variety of ways to disseminate knowledge today: mailing books, writing assignments, audio and video discs, telephone consultations, and virtually limitless possibilities of the World Wide Web. The use of distance learning methods allows to provide educational services of various kinds in remote areas, to teach without interruption from the main activity, to teach people with disabilities, to have access to educational resources of the world through the use of modern electronic communications. In today's world, education plays a key role in people's lives, which makes it necessary to constantly develop teaching methods, including distance learning. The coronavirus pandemic has revealed many shortcomings of distance education, as well as the absolute unpreparedness of students and educational institutions for this form of education. That is why it is important to develop this industry now in order to further provide comfortable education for both Ukrainian and foreign students and to ensure that in the future the education system will be ready for situations similar to the coronavirus pandemic. The article is written to assess the advantages and disadvantages of distance learning, its necessity, analysis of the requirements for distance learning materials. Prospects of technologies of multimedia representation of researched objects, processes, phenomena, adequate modeling of the subject area, and their realization.

SUPPORT OF MAKING THE DECISION ABOUT DONATION POSSIBILITY ON THE BASIS OF CIVIL LAW GROUNDS

In this paper, the actuality and importance of the process of supporting the decision-making about the possibility of donation based on the civil law grounds are proved. This realized process will help to decide about the possibility or impossibility of donation, in any case, taking into account all existing civil law grounds. A review of the literature was conducted. It proved that the known information systems in the field of donation and transplantology don’t propose the method of supporting the decision-making about the possibility of donation based on the civil law grounds. The review of the civil law grounds of donation in Ukraine was conducted. Production rules and method of supporting the decision-making about the possibility of donation based on civil law grounds has been first time developed. This method helps to the transplantation clinic staff to correctly form the Register of donors (living donors and posthumous donors), taking into account all current civil law grounds. The developed method standardizes the procedure of including donors in the Register, minimizing the impact of subjectivism and the human factor herewith. Thus, the developed method of supporting the decision-making about the possibility of donation minimizes the damage that the whole medical transplantation system may suffer and leverages the occurrence for its improvement. Experiment has shown that the developed method is workable and effective in determining the possibility of donation based on all currently existing civil law grounds in the complex. The perspective direction of the future authors' work is the design and development of the subsystem for supporting the decision-making about the possibility of donationn– on the basis of the production rules and method of support of making the decision about donation possibility on the basis of civil law grounds, which are developed in this paper.

METHOD FOR RESILIENCE FORECASTING OF THE CALAUD-ORIENTED CYBERPHYSICAL SYSTEMS

Cyberphysical system is the integration of physical space (equipment, devices and people) with computing, communication and control systems (cyberspace). The National Institute of Standards and Terminology (NIST) has defined CFS as cyberphysical systems (CFS), which are designed systems built on the continuous integration of computational algorithms and physical components. Improving the CFS provides greater opportunities for performance, adaptability, scalability, stability, security and usability, far exceeding today's simple embedded systems. CFS technology will transform the interaction of people with the engineering system [1, 2]. Today, cyberphysical systems exist in various fields, such as automotive, aerospace, civil, railway, medical. Large productions seek to increase the availability of the asset, while reducing maintenance costs through cyberphysical systems. With the development of the industry and the use of systems in production, the proposed maintenance is used constantly to avoid failures. CFS maintenance is performed when certain indicators signal that the condition of the system has deteriorated. One way to solve this problem is to provide systems with resistance properties. Such systems are able to recover quickly and continue to function in changing conditions [3-9]. Therefore, the urgent task is to develop approaches that will predict the resilience of cyberphysical systems based on cloud computing. It is necessary to build methods and tools that will monitor the condition of the CFS and predict the timely replacement of their components that may fail. The application of the process of predicting the resilience of cyberphysical systems using cloud-oriented increases the effective reliability and availability of cyberphysical systems in its life cycle by identifying future failures and reducing unscheduled maintenance. The forecasting process involves the assessment of the useful life, and the implementation of a post-forecast decision on maintenance measures in accordance with the rules. The method of predicting the resilience of cloud-oriented cyberphysical systems allows to determine the state of the cyberphysical systems using cloud computing. The mechanism used to reduce the cost of maintenance and detailed planning of maintenance operations, the apparatus of genetic algorithms.

APPROACHES TO THE CREATION OF INFORMATION AND ANALYTICAL SYSTEM FOR EVALUATION OF EFFICIENCY OF MATHEMATICAL MODELS

The article is devoted to the substantiation of the choice of the basic technique of estimation of efficiency of mathematical models which could be accepted as the basic model at the development of the corresponding information-analytical system, and also the definition of those provisions which need to be considered in the specified basic technique. The study found that as a basic method of evaluating the effectiveness of mathematical models, it is advisable to adopt a methodology based on a combination of existing approaches to evaluating efficiency, the first of which concerns the implementation of stages of evaluating the feasibility of modeling, comparative evaluation of models, the second is the quantitative assessment of significant requirements for mathematical models of phenomena, processes, and systems. These approaches should complement each other at the macro and micro levels by implementing the properties of additivity and superposition of important provisions for modeling, which are different in different approaches. In this case, the main aspects that determine the necessary areas for improvement of the basic methodology are aspects related to the creation of instrumental methods for estimating the speed and material costs of creating a model and evaluating the research result obtained using modeling, as well as substantiation of analytical dependencies. to determine the average criterion assessment of the model, integrated assessment of incremental effects, reliability of modeling, relative average values of error introduced into model calculations due to generalized consideration of parameters and factors, dependence for absolute assessment of modeling completeness.

INFORMATION TECHNOLOGY FOR COMPARING DISTANCE LEARNING SERVICES

We will remember 2020 with the coronavirus pandemic. This terrible disease has caused a long-term quarantine and therefore the whole planet, including Ukraine, has switched to distance learning. In the age of the Internet, there are many different services for distance learning, so all schools, universities, and other educational institutions have begun to use a variety of services. But after long use, it became clear that all services have their drawbacks and do not cover all the needs of teachers or students. The distance form of education opens access to education for a wide range of people who for some reason cannot study traditionally. However, in distance learning, there is a problem of lack of communication with teachers and classmates, which is the most important information channel in the learning process. Not all distance learning services are the same. An important issue for both teachers and students is the choice of a distance learning service. The article provides a brief overview of the functional features of the most popular educational tools that use the power of social media to help students in the learning process, and teachers to interact with each other, and students who are able to minimize the lack of direct communication with their social group. The article proposes an approach to the assessment of the integrated indicator, which would take into account the needs of both parties to the educational process. According to the proposed method, a comparison of the most used services that support distance learning.

ON THE NECESSITY OF IMPROVING THE METHOD OF CLUSTERIZATION OF VESSEL ROUTES AS A PROCEDURAL MODULE OF THE AUTOMATED DATA PRODUCTION PROCESSING SYSTEM IN SUN

The article is devoted to the analysis of the existing method of clustering of ship routes within the exclusive sea (economic) zone from the standpoint of substantiation of its application or identification of ways to improve it for use in the surface lighting system in the interests of national security at the state border. The study found that the existing method of clustering ship routes can not be explicitly applicable for use in the surface lighting system as a procedural module of the automated data processing system, which would provide a sufficient level of reliability to detect signs of violations of border legislation within the exclusive maritime ( economic) zone. It is also established that the improvement of the method of clustering of routes of ships should address the following issues: definition of such types of approximation for the formation of continuous routes of individual ships, which would provide a sufficient level of reliability of the results on given experimental data sets. time; justification for the choice of such a number of experimental points of location of ships and directly a combination of points that would ensure a sufficient level of reliability of the conclusions formed; study of different metrics to establish the similarity of arbitrary routes of ships from the standpoint of ensuring unambiguous conclusions; adaptation of the proposed tools to the boundary initial conditions of the studied problem; adaptation of the method for the case of arbitrary complexity of the route of vessels; formation of such a method of constructing a reference route within the desired cluster, which would ensure the reproduction of the route in the form of a continuous trajectory

TECHNIQUE FOR THE RISK ASSESSING OF THE CYBERPHYSICAL SYSTEMS’ INFORMATION SECURITY BASED ON THE VULNERABILITIES’ INTERCONNECT

Information security has been growing steadily in recent times. Every organization depends on information technology and information security of cyberphysical systems to successfully perform its work. This has become not just a condition for the stability of doing business, but the most important strategic factor for its future development, even in the current, very turbulent environment. Cyberphysical systems can contain a wide variety of entities, ranging from office networks, financial and personnel systems to highly specialized systems. The rapid development of cyber-physical systems has become due to the large number of cyberattacks, which have become one of the most powerful threats to the security of cyber-physical systems. Many studies have been conducted on the risk assessment method, and limited work has been published on quantifying the security risk of cyber-physical systems. In this paper, a technique for the risk assessing of the cyber-physical systems’ information security based on the vulnerabilities’ interconnect is proposed. Technique operates with two indicators to quantify the risk: the probability of attack success and the index of the consequences of the attack based on the graph of the vulnerability. The first indicator - the index of the probability of a successful attack is calculated taking into account the interdependencies between vulnerabilities, the second indicator when calculating the index of the consequences of the attack takes into account the impact on the physical area resulting from cyberattack. A quantitative experimental example showed whether a system risk and an optimal attack target are possible.

MATHEMATICAL MODEL OF HEAT SUPPLY INSTALLATION AS AN OBJECT OF ECONOMIC EFFICIENCY AND AUTOMATION

Solar energy is widely used in cases where the inaccessibility of other energy sources in combination with a sufficient amount of solar radiation justifies it economically. With the help of a system of solar collectors, you can significantly reduce energy costs for hot water supply and heating. In addition, the use of this system helps reduce CO2 emissions. The potential of solar energy in Ukraine is quite high. According to the climatic conditions of our region, solar activity is slightly lower than in the southern regions, so the use of flat solar collectors, which use both direct and scattered solar radiation, is effective for our region. The use of heat collectors in the home for water supply and heating of small areas is economically feasible, as it does not require significant costs. Heat supply plants using non-traditional energy sources are promising in terms of fuel savings and reduction of harmful effects on the environment. To increase the service life, economical use of heat and fuel, efficient operation of the installation requires the use of automation. The article considers the development of a mathematical model of a heat supply installation based on a solar collector. For automated control of the heat supply installation, it is necessary to know the properties of this installation as a whole and its individual elements. For this purpose, models of individual elements of the heat supply installation in a linear approximation are considered. The solar collector is conditionally divided into two parts. The model of the ground heat exchanger is developed. Also e battery models and consumer premises. The given matrices of elements of installation of heat supply are united in the uniform system of matrices. The obtained results allow estimating in advance efficiency of their use and economic attractiveness.

INFORMATION TECHNOLOGY FOR DETERMINATION OF CHARACTERISTICS ON THE FACE FOR EMOTIONAL RECOGNITION

One of the ways to process an image presented in the form of a set of pixels, in order to further identify, classify the objects present on it is to display the specified set in the form of sets of certain features. Such features are not universal in nature, but rather significantly depend on the tasks under consideration. For certain classes of problems, such features (model) are selected that best allow the application of appropriate methods to solve the problem. The paper considers a class of problems for recognizing the emotional state on a person's face. In, a convolutional neural network (CNN) is used to detect emotions. CNN differs from multilayer perceptron (MLP) in that they have hidden layers called convolutional layers. The proposed method is based on a two-tier CNN system. At the first level, the background of the image is removed to better reflect emotions. A standard CNN network module is used to obtain the primary expression vector (EV). EV is formed by tracking the relevant important points of the face. EV is directly related to changes in facial expression. EV is obtained using a basic perceptron unit plotted on a face image with the background removed. In the proposed model at the last stage, there is a non-evolutionary perceptron layer. Each of the convoluted layers receives input data (images), converts them, and then takes them to the next level. After detecting a face, the CNN filter of the second part captures parts of the face, such as eyes, ears, lips, nose, and cheeks. The authors agree that the method has some limitations, and especially requires high computing power when setting up CNN. The technology of determination of characteristic features on the face for recognition of emotional manifestations is presented and experimentally investigated.