ABSTRACT.  

Automated Machine Learning, or AutoML, is a newly emerging field in Machine Learning. It promises to automate predictive modeling, democratize machine learning to non-experts, boost the productivity of experts, ensure the statistical validity of the modeling process, and even surpass human experts in quality. AutoML should not only strive to produce a high-quality model, but all information, explanations, interpretations, and decision support a human expert would.  In this talk, we’ll present the challenges of AutoML and the design choices we made to construct the Just Add Data Bio, or JADBio for short, AutoML platform. We will also present industrial and health applications of JADBio.

BIOGRAPHY.

Ioannis Tsamardinos, Ph.D., is a Professor at the Computer Science Department of the University of Crete, CEO, and co-founder of JADBio, a University start-up. He obtained his Ph.D. from the Intelligent Systems Program at the University of Pittsburgh in 2001. Prof. Tsamardinos’ main research directions include machine learning, bioinformatics, and artificial intelligence. More specifically his computer science work emphasizes automated machine learning, feature selection, and causal discovery. Prof. Tsamardinos has over 140 publications in international journals, conferences, and books. Distinctions with colleagues and students a Gold Medal in the Student Paper Competition in MEDINFO 2004, the Outstanding Student Paper Award in AIPS 2000, the NASA Group Achievement Award for participation in the Remote Agent team, and others. Statistics on recognition of work include more than 10000 citations (1000+ a year), and an h-index of 40 (as estimated by Google Scholar). Ioannis has been awarded the European and Greek national grants of excellence, the ERC Consolidator, and the ARISTEIA II grants respectively.

ABSTRACT. The adjustment of the previously developed algebra-algorithmic tools towards the automated design and generation of programs that use neuroevolutionary algorithms is proposed. Neuroevolution is a set of machine learning techniques that apply evolutionary algorithms to facilitate the solving of complex tasks, such as games, robotics, and simulation of natural processes. The developed program design toolkit provides automated construction of high-level algorithm specifications represented in Glushkov’s system of algorithmic algebra and synthesis of corresponding programs based on implementation templates in a target programming language. The adjustment of the toolkit for designing neuroevolutionary algorithms consists in adding the descriptions and software implementations of the relevant elementary operators and predicates to a database of the toolkit. The use of the toolkit is illustrated by an example of designing and generating a program for the single-pole balancing problem, which applies the neuroevolutionary algorithm of the NEAT-Python library. The results of the experiment consisting in the execution of the program generated with the algebra-algorithmic toolkit are given.

ABSTRACT. It is well known that the functional programming paradigm provides certain advantages, namely increasing the reliability of software development, improving software verification and validation, providing software artefact reusability, and scaling software solutions. But this paradigm is not free from disadvantages also. The most serious of them are redundantly memory usage and low-performance productivity. The ideas concerning the implementation of randomization mechanisms into reduction strategies, which is a significant part of the kernel of functional computing systems, are considered in the paper. The authors present simulation tools for studying such a mechanism and give their grounding. They also present the experimental results of the comparison analysis of variants reduction randomization.

ABSTRACT. This paper focuses on the application of reinforcement learning methods to solve the game of Sokoban. While the game is relatively easy for humans to solve, it poses a significant challenge for computer algorithms due to the irreversible nature of specific moves. To predict which actions will lead to such undesirable states is often difficult for a learning agent - a common problem in tasks requiring planning. To address this issue, we propose using a Monte-Carlo tree search (MCTS) algorithm and a heuristic CNN specially trained to distinguish between undesirable, neutral, and desired game states. We experimented with different heuristic variations used to solve the game and compared them against each other. We employed MCTS in two different setups: one with a CNN trained using data obtained during the solving process and one without such training. We also varied the number of rollouts for each move in MCTS and compared the results. The general aim is to improve the performance of learning agents in tasks that require planning to avoid unwanted states.

ABSTRACT. The article discusses the issue of pre-service professional training of preschool teachers on the example of the course Workshop on Children’s Play Activities for students of the first (Bachelor's) level of higher education of the educational and professional program 012.00.01 Preschool Education. This paper focuses on the ways of developing the professional competence of Bachelors of Preschool Education through online interaction. The place of the course Workshop on Children’s Play Activities in the e-learning system of Borys Grinchenko Kyiv University is presented. The article describes the structure of the electronic online training course, the logic of thematic unit mapping of the course. Authors suggest effective services and resources for online interaction of participants of the educational process, namely: during lectures (Google Meet, Google Chat, Google Hangouts, Google Classroom, Webex, Zoom, Lucidspark, Agile, Canva, Mentimeter, Kahoot, Piktochart); seminar sessions (Hangouts, Zoom, Easel.ly Padlet, Piktochart, AnswerGarden, Slides, Prezi, Learningapps, Socratic); and practical classes (Microsoft Teams, Migo, Whiteboard, Agile Workflows, Class Dojo, Mural, GoAnimate, GIFAnimate, Slasstools, Learningapps, Triventy, Yumpu, Emaze, Genially, Flippity, Wordwall, Kahoot). Authors claim that usage of information and communication technologies is effective for facilitating online interaction, increasing students' engagement with theoretical material, and making the preparation to seminars and practical classes easier for students. Using the suggested services and resources will allow the organization of dynamic and engaging classes, as well as promoting active interaction with students even in online setting.

ABSTRACT. The problem of designing didactic processes has not been solved to this day, although some specific issues are considered. The article presents a generalized approach model based on the reference ADDIE model including the organisational level to the design and implementation of the didactic process. The model uses Bloom's taxonomy and Gardner's Theory of Multiple Intelligences. The model has been modified towards a dynamic design oriented on quality, efficiency and adaptive learning. A competency-based decomposition is used. The lowest level of decomposition is the activities level where the information flows as well as the learning and forgetting are taken into account. The competences-objectives-activities-data linking is discussed. The level of activities is represented in the form of an electrical-like network. The network represents differential equations describing dynamic learning and forgetting as well as the structure and information flows of the didactic process. The element models can be easily extended. The networks can be simulated and optimized. The results can be used during the design process. The simulation result as well as the structure of network equations enables inference. The article presents the results of simulations of some aspects of didactic processes including scheduling and sequencing, gap effect, workload, predicting achievements and effectiveness.

ABSTRACT. The article highlights the theoretical foundations of measuring and interpreting student attention levels, engagement, and feedback, an attempt is made to characterize the possibilities of using automated software products for this purpose. The article identifies the psychological characteristics of students of the Zoomers generation, which affect the educational process and cause the revision of educational technologies. A comparative analysis of traditional and distance learning was carried out. Based on the characteristics and interpretation of visual markers for measuring the attention and involvement of the acquirers, the requirements for the software product for determining the degree of involvement and the degree of information perception are described. The article discusses the main AI-based tools that allow you to assess the level of audience attention in real time during online conferences (EmotionCues, HEADROOM and MeetingPulse) from the standpoint of the effectiveness of the analysis of the emotional component and the possibility of integration with online conference services. The theoretical analysis is the basis for further research on the measurement and interpretation of students' attention, engagement and feedback.

ABSTRACT. An integral part of the global task of increasing the efficiency of movement processes and performing specified technological operations by a mobile robot is the choice of a movement path from the initial (starting) to the final (target) point. At the same time, a preliminary assessment of the proposed route of movement is an important stage in the development of a program for its movement. In this paper, we consider the problem of finding the shortest path for moving the mobile robot in a multilevel environment, taking into account the features of transitions between levels, restrictions, the presence or absence of obstacles along the way. An information system for calculating the shortest route from the starting point to the target point is proposed based on the NavMesh navigation system in the Unity virtual environment. The results of computer simulation of robot movement in a multilevel environment show that the system successfully finds the shortest route for various scenarios.

ABSTRACT. A drone is a complex hardware (HW) and software (SW) system that uses various wireless data transmission technologies (Wi-Fi, LTE, 5G, Bluetooth, etc.). To transfer data, it uses various communication protocols: both specific and generally known. Drones can use digital protocols and embedded systems such as I2C, SPI, Serial, CAN, 1-wire, etc. The flight controllers used in the drone: popular microcontrollers Atmega, STM (ARM), Intel Movidius, sometimes low voltage Intel/AMD x86/x64. Drones use communication protocols, operating systems (OS) in the control device, which are similar to other digital systems. Vulnerabilities of the drone's main subsystems (both HW and Sw) can be used by attackers to carry out cyber-attacks on them. Drones can perform complex tasks, and some cyber-attacks (for example, Denial-of-Services - DoS) can lead to the failure of individual components of the drone and the entire system. Guidelines for protecting against drone attacks are provided by many organizations that develop cyber-security standards (NIST, CERT, CISA, etc.). Measures to prevent cyber-attacks can be applied to drones, with some adjustments to their parameters and architectural features. There are also recommendations for protecting drone components and also methods of communication protocols protection from cyber-attacks. The authors of this study offer a comprehensive approach to the analysis of vulnerabilities of drone subsystems, which includes a system analysis of drone architecture, vulnerability analysis by different vulnerability databases, and their systematization. The purpose of this study: analysis and systematization of the most serious vulnerabilities of drone subsystems and cyber-attacks that can affect them, and recommendations for their prevention, detection and mitigation.

ABSTRACT. The work is a continuation of the research devoted to the development of a multidimensional Kalman filter connected at the output of the built-in helicopter’s turboshaft engines mathematical dynamic model to improve the accuracy of helicopters turboshaft engines parameters identification and achieve high quality automatic control. The main difference is the use of radial basis functions neural networks, in which the multivariate Kalman filter is a training algorithm. The work illustrates well-known mathematical expressions underlying of optimal multidimensional filtering algorithms. The methods of mathematical modeling in the Matlab environment tested the proposed algorithms. The simulation results showed that the use of neural networks trained by the multidimensional Kalman matrix filter as part of the model of helicopters turboshaft engines built into the automatic control system allows achieving high indicators of the accuracy of identifying the parameters of helicopters turboshaft engines automatic control system – up to 0.9975, then in practical analogues

ABSTRACT. The relevance of this work lies in the need for qualitative evaluation of innovative projects, which are characterized by various technical and economic indicators, and, as a rule, significant capital investments are required for their implementation. The use of multi-criteria methods, in particular the modified Fuzzy TOPSIS, will allow to increase the accuracy of decision-making when choosing the optimal innovative project. The purpose of the work is to select the optimal innovative project through the use of the developed software application based on the modified Fuzzy TOPSIS method, as well as to investigate the influence of the parameters of the specified method on the decision-making result. To modify the Fuzzy TOPSIS method, the Euclidean distance, the Hamming distance, and the Chebyshev distance were used to determine the proximity coefficients. Another element of the modification of the Fuzzy TOPSIS method was the ability to calculate results using various fuzzy numbers, such as triangular, trapezoidal, and Gaussian. A comparative analysis of the obtained results with other multi-criteria decision-making methods was carried out. A data entry template for experts has also been created and the results presented in a clear and easy-to-read format

ABSTRACT. This paper focuses on developing a flexible and scalable model for effectively using fog resources for sharing data processing capabilities in a model with both local and cloud components. The Node-RED was chosen as the best tool to visualise the authors' algorithms and approach. The literature review on this direction shows that there are no works that describe the use of fog resources for data processing in systems with both local and cloud components using some regulation factors. A solution to this problem was proposed in this work as working hypothesis. Therefore, authors think that the proposed solution is novel and has scientific value. To prove this hypothesis, authors have prepared the setup using Node-RED and conducted the corresponding experiments for the future. Adding an extra feedback loop between services in the cloud and in the local network is proposed to control the number of additional processing nodes. The corresponding setup has been prepared to investigate this approach. It includes the cloud part with processing service, extra nodes regulator, and the local domain that contains the primary node, additional nodes, load balancer, and the sources of data streams. The feedback loop is organised as a data stream, flowing from the cloud to the local system. Such an approach reduces latency caused by data processing at nodes in the cloud. Also, it allows reusing free fog node resources. The work also suggests further research directions and system improvements.

ABSTRACT. Modernization of the system of the property registers and cadastres is an essential component for the development of society, automation of services providing, optimization of register’s operating costs, efficient operation of subsidy and tax systems, and reduction of property fraud. An analysis of the current state of the property public registers system has been given. It has been established that considering Ukraine’s historical development, it is expedient to modernize the existing information systems and ensure information interaction between the existing registers and cadastres, but not develop the new system (property cadastre). The interoperability of the data in property information systems has been described. The functions of the main stakeholders on property registration (land surveyors, a state employee of the state architectural service, local government architects and private engineers, and state registrars of property rights) have been described in the form of a case UML diagram.

ABSTRACT. The article updates the problem of the application of digital technologies in various professional methods and their integration in the educational process of institutions of higher pedagogical education. According to the authors, the formation of subject-methodical competence of future primary school teachers is not possible without the formation of abilities: navigate in the information space, use existing and, if necessary, create new electronic resources; use modern digital technologies in the educational process. For the formation of professional skills and abilities in students of pedagogical specialties, the quality of educational content of a methodological nature, created by means of digital technologies, is essential. In accordance with the purpose of digital technologies and their application in professional teaching methods, the most effective sets of programs and applications were determined for students to create educational and methodological support for the educational process in primary school. Such as: mind maps (summary diagram of interactive lessons),presentations (demonstration of educational materials), comics (visualization of educational information in frames-pictures), word clouds (visualization of educational content in keywords), infographics (visualization and structuring of a large amount of educational information), virtual boards (coordination of work in groups in class, during the web quests, organization of student communication), interactive tasks and online tests (creation of exercises to acquire skills, development of tasks and tests to reveal knowledge and skills).

ABSTRACT. Recent years of the covid pandemic and military threats have significantly exacer-bated the need for young people to obtain university education remotely. This al-so increases the requirements for the effectiveness of using the didactic capabili-ties of digital means and tools of distance education, and the quality of relevant content for organizing and supporting students' independent educational activities. The value of independently acquired, built, and not obtained in the finished form of knowledge increases. Able to effective independent cognitive activity can only be a person who has the appropriate cognitive skills, and is aware of his own cognitive processes. Therefore, filling students' independent work with metacog-nitive content should be considered as the subject and purpose of training, along with the traditional goal of the university, which is the formation of subject knowledge and skills. The subject of our research is the use of means and tools of LMS Moodle for the organization of independent work of students based on a metacognitive approach in the study of physical and mathematical disciplines. The article analyzes the problems of organizing students' independent work and offers some ways to solve them using the Moodle functionality. Examples of the use of individual Moodle activities by the article's authors for the organization and sup-port of independent work of students are given.