ABSTRACT. This is a welcome talk from the Conreferece Program Chairs. It will provide the useful information for the participants on the conference program, social events, etc.

ABSTRACT. In this talk, we delve into the cutting-edge realm of Neuro-Symbolic Artifi-cial Intelligence (AI), focusing on the synergistic integration of Knowledge Graphs and Generative AI such as Large Language Models. Neuro-Symbolic AI represents a transformative approach that combines the robust, interpreta-ble reasoning capabilities of symbolic AI with the adaptive, data-driven strengths of neural networks. The talk will illuminate how this fusion offers a promising pathway towards more intelligent, explainable, and reliable AI sys-tems. As a showcase of our approach towards neuro-symbolic AI we will demonstrate Corporate Memory, an enterprise ready Knowledge Graph and Neuro-Symbolic AI platform used by major Enterprises as well as the Open Research Knowledge Graph. The ORKG is representing research contribu-tions in a structured and semantic way as a knowledge graph. The advantage is that information represented in a knowledge graph is readable by machines and humans. For creating the knowledge graph representation, we rely on a mixture of manual (crowd/expert sourcing) and (semi-)automated techniques leveraging Large Language Models. Only with such a combination of human and machine intelligence, we can achieve the required quality of the repre-sentation to allow for novel exploration and assistance services for enterpris-es and researchers. As a result, a scholarly knowledge graph such as the ORKG can be used to give a condensed overview on the state-of-the-art ad-dressing a particular research quest, for example as a tabular comparison of contributions according to various characteristics of the approaches.

ABSTRACT. In this article, we describe the process of generating smart contract code using AI-driven tools and its subsequent verification via the insertion modeling system. Initially, we outline the process of defining and specifying smart contract requirements, the creation of acceptance criteria in the Gherkin language. Next, we generated the smart contract code in the Solidity language based on delineated requirements using AI. The contract`s purpose is to manage the token unlocking processes for decentralized platform investors. We describe the process of translating the smart contract code into the algebraic specifications of the IMS system through ANTLR grammar. To ensure the functional correctness of smart contract operations, we formalized the acceptance criteria described in the goal state syntax and checked whether the smart contract code reaches the goal state defined in the requirements specification. The study shows how these techniques can improve the quality and reliability of smart contracts in distributed systems. The described approach allows us to check the correctness of the smart contract’s functioning logic and its compliance with the given specifications.

ABSTRACT. Climate factors play a decisive role in crop yield fluctuations. From the perspective of wheat cultivation in Ukraine, three agroclimatic zones can be distinguished. This study investigates the impact of climatic factors on wheat yield fluctuations for each zone. It is demonstrated that the influence of nonlinear climate factors significantly enhances the accuracy of wheat yield prediction. Using machine learning techniques, models are constructed to forecast future yields with a prediction horizon of 3 months. Such forecasts can provide a basis for optimal investment and marketing decisions in the grain market. The proposed methodology can be applied to forecast yields of other agricultural crops as well.

ABSTRACT. An analysis of the principles and methods of using message brokers in the process of determining the configuration and adaptation of the software was carried out. It has been established that message broker technologies allow to distribute the load on the resource by using task queues. The types of configuration of message brokers were studied and it was established that for the effective implementation of the adaptation process it is advisable to use the dead-letter exchange queue type. A messaging model is designed to enable dynamic determination of software configuration. It has been established that the use of message brokers when adapting software makes it possible to distribute the process of determining the configuration for software systems. An algorithm for software adaptation based on the use of a combined architecture and message brokers is proposed. It has been established that the proposed solution allows to reduce the load on the resources of the web adaptation service.

ABSTRACT. We present enhancements to the reinforcement learning (RL) approach used in the NeurIPS 2023 HomeRobot: Open Vocabulary Mobile Manipulation (OVMM) Challenge, focusing on augmenting the baseline model with advanced semantic segmentation and skill policy modifications. More specifically, we introduce refined semantic segmentation model ( integrating the YOLOv8 and MobileSAM), improved place skill policy and a high-level heuristic strategy, which collectively advance the overall success rate from 0.8 to 5.2 (+550% relative) and the partial success rate from 9.7 to 25.8 (+165% relative) on the Test Standard split of the challenge dataset (ranked 2nd on the public leaderbord). These enhancements enabled our agent to achieve 3rd place in both the simulated and real-world stages of the competition. This paper details the strategies employed, discusses the insights gained, particularly in semantic segmentation and skill-specific training, and outlines potential avenues for future enhancements in embodied AI systems within open-vocabulary contexts.

ABSTRACT. Alzheimer's disease is a neurodegenerative disorder characterized by a progressive decline in cognitive and memory functions, as well as the ability to carry out basic activities. Symptoms typically manifest in later years, affecting individuals of all age groups, particularly older adults. Timely identification is essential for accurate diagnosis of this condition. Manual diagnosis by healthcare professionals is often time-consuming and prone to errors due to the widespread prevalence of the disease. Recent advancements in information technology, including deep learning, machine learning, and artificial intelligence, have enabled the development of autonomous systems that require minimal human intervention. This study utilizes deep learning techniques in combination with image processing methods for the detection of Alzheimer's disease. A novel approach to treating Alzheimer's disease is proposed and evaluated using a real dataset obtained from the ADNI repository on Kaggle. The dataset classifies images into five distinct categories: cognitive normal, mild cognitive impairment, early mild cognitive impairment, late mild cognitive impairment, and Alzheimer's disease. Three deep learning models, namely DenseNet, Inception V3, and ResNet 50, are implemented and assessed. The results indicate that the proposed models effectively identify Alzheimer's disease at all stages. Specifically, ResNet 50 demonstrates superior accuracy in detecting Alzheimer's disease and cognitive normal classes, while DenseNet exhibits stronger performance in detecting mild cognitive impairment in its early and late stages.

ABSTRACT. Artificial intelligence approaches, particularly neural network technology, play an increasingly important role in modern systems for detecting and preventing intrusions in computer systems and networks. Existing classification models reveal system’s anomalous behavior and the type of intrusion to which an attacked system is exposed. However, neural network technology has several problems, such as false positive detections, adversarial attacks, heterogeneity of training sets, and the inability to classify intrusions outside of the training dataset. The use of algebraic methods and modern solver systems for the accurate detection of real-time attacks is not common, as they are much slower than neural network classification. This work uses formal constructs, such as cognitive networks, to leverage the synergy between neural network technology and algebraic methods. These constructs are a composition of a neural network and transition system networks based on algebraic behavioral models. The use of algebraic modeling technologies and neural networks is demonstrated with examples that present the real-time prevention of attacks in software and network environments. This paper demonstrates the use of a cognitive network that combines these two techniques as a novel approach for model training and cyberattack detection.

ABSTRACT. A description of the typical structure and typology of simulation models, which are used for educational and research purposes in population ecology at V. N. Karazin Kharkiv National University. These models decide a certain set of reasons sufficient to explain the appearance of a certain feature of the studied system, and conditions do these reasons ensure the appearance of this feature. Type I models determine the dynamics of a certain process, which occurs under certain initial conditions, parameters and conditions of the experiment. Type II models establish the probability distribution of simulation results over a series of iterations under identical initial conditions. Type III models determine the influence of various combinations of initial parameters of simulation modeling on the probability distribution of its results. Type IV models repeat and store in the form of a multidimensional array the results of simulations with different combinations of key parameters, which allows you to establish which of these combinations correspond to certain simulation results. Using the R programming language, we created the Simpson's Paradox model as an example of the types of models considered. This model explains the phenomenon of the expansion of altruistic behavior and helps identify the specific conditions under which altruists can succeed in a hypothetical population. It has been shown that in many cases group selection in a subdivided population can outperform individual selection. The authors believe that the methods used to build the discussed models can be useful for solving a number of other research problems.

ABSTRACT. The work is devoted to the review and analysis of software for designing control programs and 2D/3D modeling of the mobile robots for academic and industrial applications. A generalized structural diagram of the hierarchical mobile robot control system with decentralized software processing of information and separate software and hardware components located remote from each other are presented in the paper. Various programming environments which represent a wide range of tools for creating different types models and systems are considered for building the mobile robots and their control systems. Moreover, the issues of using graphic and text software environments with high-level programming languages have been reviewed. LabView, EV3-G, TRIK Studio, ROBO Pro, Arduino and Scratch were considered among the integrated development environments. Player/Stage, Gazebo, RoboDK and others were considered among the software packages for modeling the operation of the mobile robots. Presented examples of academic and industrial software are analyzed according to the following criteria: mathematical expressions, computational models, interpretation, standalone use, code generation, debugging, tutorials, free, platforms, constructors and development prospects. And also robotic systems simulators are considered as tools for testing the efficiency, safety and reliability of control algorithms through 2D and 3D simulations.

ABSTRACT. The article covers the phenomenon of accessibility in the IT industry, as well as technical and organizational ways of its implementation in modern commercial web development projects. The research also describes a general context of accessibility and defines its in the ICT sphere, provides an overview of the basic requirements, prerequisites, standards, and practices for ensuring accessibility, which may be applicable in the real-life industry cases. Special emphasis in the article is placed on the automated of testing, as a modern trend in quality assurance, which involves the implementation of autonomous checks of functional and non-functional requirements to the product through the use of specialized tools and approaches. The stages of implementation of automated testing on projects are revealed in the article, a general classification of tools and approaches is provided. During the analysis of practices and industrial experience of implementation of automated accessibility testing, its features and limitations were identified in the conditions of the commercial development process in general, as well as in its individual phases, such as analysis of requirements, creation of designs, writing the software code and testing. The work provides a classification of specific tools of web accessibility automation by their levels and nuances of implementation in accordance with the previously defined project stages and the general process of continuous accessibility, as well as the analysis of the most popular and representative software of each category. It was concluded that the comprehensive and systematic use of automated testing in the context of ensuring accessibility is one of the key factors for the success of implementing the principles of inclusion in modern software development processes and products.

ABSTRACT. Currently, image display systems utilizing four subpixels – white, red, green, and blue – are garnering significant interest. This study explores the potential to enhance these systems by expanding their dynamic range, improving quality, and increasing energy efficiency. The research employs Huebel's physiological model of color vision and proposes the WRGB color model, which includes the use of a white subpixel. To compare images reproduced by the classic RGB model and the LED WRGB model, a program was developed. This program performs color conversion from the RGB model to LED WRGB, enabling a comparison of the energy consumption of LEDs and the spectral characteristics for color reproduc-tion in both models.

ABSTRACT. This paper describes the methods for normalizing Ukrainian texts which contain numbers in digital form. The proposed application converts numerals into letter form with the corresponding morphological features. Ukrainian natural language processing tools are used to determine the morphological attributes of numerals. The conversion of texts considers the rules of Ukrainian grammar. The accuracy of the developed methods reaches 94.17 %. The final software product will serve as a tool for pre-processing sentences with numerals for Ukrainian speech synthesis.

ABSTRACT. The retrieval-augmented generation (RAG) enables retrieval of relevant information from an external knowledge source and allows large language models (LLMs) to answer queries over previously un- seen document collections. However, it was demonstrated that tradi- tional RAG applications perform poorly in answering multi-hop ques- tions, which require retrieving and reasoning over multiple elements of supporting evidence. We introduce a new method called Multi-Meta- RAG, which uses database filtering with LLM-extracted metadata to im- prove the RAG selection of the relevant documents from various sources, relevant to the question. While database filtering is specific to a set of questions from a particular domain and format, we found out that Multi-Meta-RAG greatly improves the results on the MultiHop-RAG benchmark. The code is available on https://github.com/mxpoliakov/Multi-Meta-RAG.

ABSTRACT. The increasing prevalence of artificial intelligence (AI) and machine learning (ML) in various sectors has led to a growing need for higher education institutions (HEIs) to adopt data-driven decision making (DDDM) processes. This study explores the use of ML techniques to identify the target group of applicants, enabling the effective allocation of resources for marketing and careers activities. The research highlights the importance of access to diverse and large datasets in order to train accurate ML models. HEIs with established AI teams, training strategies, collaborations with AI service providers, and a digitised and robust data infrastructure are better placed to make effective use of AI/ML tools. For higher education authorities, it is crucial to interpret the insights derived from applicant data. Decision support methods using AI include expert systems, machine learning, neural networks and deep learning architectures. ML can improve various areas within higher education institutions, such as predicting applicant numbers, personalizing education, preventing dropouts, improving efficiency, recruiting and automating routine tasks. The aim of this research is to develop models based on machine learning that can accurately predict the probability of an applicant's admission to an HEI using DDDM. Among all the methods, the KNN algorithm showed the best result in predicting the admission of applicants with an accuracy of 0.8378. The logistic model also has a high accuracy of 0.8108. The KNN model is the best according to the RMSE criterion. The research provides insights into the use of ML techniques for data-driven decision making in higher education, while emphasizing the need for public oversight, stakeholder involvement and balanced integration of ML into the educational process.

ABSTRACT. In today's rapidly evolving business landscape, the pace of change is driven by disruptive technologies and shifting consumer preferences. One model gaining significant traction, particularly in web services and software for product IT companies, is the Freemium pricing strategy. The Freemium model provides basic services for free while charging for additional premium features. Researching the Freemium business model is highly relevant as it presents opportunities for enterprises to attract users and monetize their product or service offerings. However, maximizing this model's effectiveness requires a deep understanding of user acquisition dynamics, preferences, and willingness to pay - factors that can be experimentally modeled using Python functionality. Goal of paper is to determine the optimal premium price point and factors influencing revenue under the Freemium model for IT companies using the Python programming language. We identified the optimal premium subscription price for maximizing IT company profits under the Freemium business model through Python simulation experiments. Our methods include optimization models (profit maximization in discrete Freemium models), simulation modeling (discrete pricing model), and graphical methods (interpreting economic metrics under premium user share impact) in Google Colab. In the paper a Freemium pricing model for premium customers was investigated, visualized, and experimentally generated using Python, aimed at maximizing the profits of IT companies through simulation modeling using Python. A discrete dynamic model was developed relating profit to the share of premium users. The optimal premium subscription price for maximizing IT company profits was determined in Python through simulation experiments in Colab Environment.

ABSTRACT. We present an optimization study of the Vision-Language Frontier Maps (VLFM) applied to the Object Goal Navigation task in robotics. Our work evaluates the efficiency and performance of various vision-language models, object detectors, segmentation models, and multi-modal comprehension and Visual Question Answering modules. Using the val-mini and val splits of Habitat-Matterport 3D dataset, we conduct experiments on a desktop with limited VRAM. We propose a solution that achieves a higher success rate (+1.55%) improving over the VLFM BLIP-2 baseline without substantial success-weighted path length loss while requiring 2.3 times less video memory. Our findings provide insights into balancing model performance and computational efficiency, suggesting effective deployment strategies for resource-limited environments.

ABSTRACT. Without a doubt, technology has transformed the educational process. Massive availability of the internet and cheap personal devices (mobile phones and laptops) allows more people to participate in the education process, as never before. New set of technologies such as VR, AR, and AI together with the massive popularity of video games and digital entertainment open new set of possibilities for educators and online course designers. Student-to-online course interaction is a vital part of the online educational framework. There are three phases of student-online course interaction ranging from personal educational goals to direct and indirect student-service interaction, and various gamification techniques. The Orthodox game design and development methodologies have been used to build complex web service frameworks. This framework consists of several modules and parts, while the stu-dent-system interaction is expected to happen in cycles (iteration loop). During the study-interaction process and in the results of course/material completion students are awarded and provided with automated support, the next step of content. The Gamification framework is a semi-automated system that must be connected to an existing online educational platform.

ABSTRACT. The article examines the role of digital tools to support formative assessment in university mathematics teaching. Formative assessment is considered as a pur-poseful systematic process of collecting, analyzing, and using information about the educational achievements of students to improve their learning and improve teaching. Considering the peculiarities of mathematical disciplines that distinguish them from other academic disciplines, the article argues the importance of forma-tive assessment in mathematics and formulates key indicators of mathematical competence that are subject to assessment. Pedagogical tools of formative as-sessment in mathematical disciplines are proposed and the ability of digital tools to increase their effectiveness is argued. Examples of the use of individual tools from the teaching practice of the authors of the article are given.

ABSTRACT. Methods of constructing jamming-tolerant codes based on multi-position combinatorial configurations with a ring structure of the type of ideal ring-bundles are investigated to create coding systems that detect and correct errors, with improved quality indicators in terms of power and jamming. A general description of the method of building a multi-position jam-resistant code based on ideal ring- bundles, and methods of increasing the power of multi-position jam-resistant ideal ringbundles codes is given. A comparative analysis of the parameters of the proposed code with the parameters of well-known codes was carried out. A study of the impact on the correction data was conducted depending on the number of errors and how many errors my program corrects. A method of simplified synthesis of an error-resistant multi-position code sequence based on ideal ring bundles has been developed, and an effective algorithm for encoding and decoding information has been created. The developed software application allows error-correcting coding correction based on ideal ring bundles (correction of up to 25% of the code length and detection of up to 50% of the code length).