Andrii Khudyi

Automation of database indexing is a crucial component of modern database management systems that enhances search performance, scalability, and relevance in large-scale data environments. This paper explores the application of deep learning algorithms for building and optimizing vector indexes capable of automatic adaptation to changes in data structure and query patterns. An experimental comparison was conducted between traditional indexing methods (B-Tree, GIN in PostgreSQL) and vector-based indexing using Sentence-BERT embeddings implemented in FAISS and Milvus systems. The study involved constructing a corpus of 1–5 million text records and performing a series of tests to evaluate performance metrics, including average query time, latency (95th percentile), index size, memory usage, and the search completeness metric (recall@10) relative to the BM25 baseline search. The results revealed that vector indexing achieved up to 0.94 recall@10 with comparable query latency, significantly outperforming traditional approaches in semantic search efficiency. The practical contribution of this work lies in demonstrating the feasibility of hybrid DBMS architectures that combine traditional SQL indexing for fast filtering with vector-based methods for semantic refinement. This approach lays the foundation for the development of intelligent, self-optimizing databases capable of autonomous adaptation. Future research should focus on reinforcement learning for dynamic index restructuring, optimization of computational costs, and the integration of vector indexing into cloud-based and distributed systems.

This paper presents an adaptive approach to foreign-language learning in a mobile application based on spaced-repetition algorithms. The approach implements principles of the Ebbinghaus forgetting curve and a spaced repetition system (SRS) to dynamically schedule presentations of learning material. Flashcards provide a multimodal representation of vocabulary items - textual form, transcription, an in- sentence example, and audio accompaniment for training listening and pronunciation. The translation is revealed only after a tap, which promotes active recall; knowledge is reported via gestures: swipe right - known, swipe left - unknown, swipe up - learned (the card is archived and won’t reappear). The mechanism aligns with SM-2-style modeling: intervals automatically increase for well-learned items and shrink for difficult ones, keeping the system focused on “weak” words and avoiding excessive repetition of familiar material. The algorithmic integration is specified: the data model stores the last review date, interval, repetition count, knowledge rating, and an ease factor, enabling personalization of the schedule to individual memory characteristics. Interface decisions combining gesture interaction with audio accompaniment are demonstrated, and the expected effects are outlined: longer-term retention of vocabulary, fewer redundant exposures, and stronger motivation due to transparent progress feedback. The approach is scalable and suitable for further extensions (e.g., parameter tuning or data-driven models for scheduling the next review) without requiring complex infrastructure on mobile platforms.

This article presents an integrated Advanced Driver Assistance System (ADAS) that combines several key functional modules, such as collision warning, lane detection, traffic sign recognition, and pothole detection, which are implemented using modern deep learning models, particularly YOLOv8n. The system is optimized for devices with limited computational resources, such as Raspberry Pi or NVIDIA Jetson Nano, by employing a modular architecture and parallel data processing to ensure realtime performance. This research provides an overview of existing ADAS solutions and proposes new approaches that significantly enhance the efficiency of such systems. Key innovations include an efficient approach to lane detection based on object detection models, real-time traffic sign recognition with a flexible extraction and classification process, and a novel pothole detection system optimized for dashcam recordings. Additionally, the proposed driver alert system, which uses an LED strip, allows for intuitive hazard awareness without distracting the driver. Preliminary results confirm satisfactory detection accuracy across all components, although further optimization is required for successful deployment on low-resource devices.

The need to enhance query performance in databases within the contemporary information environment is crucial for ensuring efficient operations across various domains. This paper is dedicated to an analysis of database indexing techniques aimed at understanding their impact on query performance and efficiency. It meticulously examines various types of indexes, including B-trees, hash tables, and textual indexes, analyzing their advantages and limitations. The investigation into the influence of these techniques on query execution time reveals that performance depends on various factors such as query complexity and data volume. Special attention is given to selecting the optimal index type based on specific database needs and characteristics. The research underscores the importance of considering data volume and database structure in choosing effective indexing methods. Additionally, the advantages and disadvantages of different indexing techniques are scrutinized, taking into account their impact on query execution speed and efficiency. Of the study indicate that the correct selection and utilization of appropriate indexing strategies can significantly enhance the performance of database management systems, providing swift and efficient access to information for users. Ultimately, this work makes a significant contribution to understanding and practically applying indexing techniques to improve query performance in databases.

Phasor diagrams are a powerful tool for visualizing and understanding the distribution of current, voltage, and power in electrical systems. During Russia's war against Ukraine, our energy industry has become very vulnerable to enemy attacks, and therefore needs a quick and effective recovery. Energy specialists lack software tools for working with the power system, and in the period of development of artificial intelligence, creating such tools is not so difficult. For example, familiar specialists often have to build vector diagrams - tools for visualizing and understanding the distribution of current, voltage and power in electrical systems. Using a combination of frameworks for working with artificial intelligence and creating a graphic shell, you can achieve the desired result in a few months, and at the same time do a useful thing for our victory. The YOLO model (based on the PyTorch framework) and the QT framework are considered among the proposed tools for creating an intelligent vector diagram construction system. The role of artificial intelligence is to recognize electrical elements in circuits and their connections to each other. Creating a user interface is an equally important thing, which is already implemented in many places with the help of QT. As of today, there is no specialized software tool for solving the problems of manual construction of vector diagrams, but the proposed approaches are already used for its creation. The system uses the Yolov5 model to recognize electrical elements on the circuit. The model is trained on more than 150 images and is able to recognize hand-drawn diagrams. Recognition is run as a separate process from the main program written in C++. This part of the system processes the input data from Yolo, saves it in a convenient format, creates a user interface, and displays the result as a vector diagram.

In this paper, the problem of precise determination of the constant value of the Hurst fractal index for the time series is considered. In the classical representation, the value of the Hurst index is neglected by the constant, which is mostly equal 0.5 or, in general, includes it in the variable. In addition, the existing methods for determining this indicator are empirical. In the proposed method, the Hurst index was measured through a predetermined value of the fractal dimension. As a result, another indicator, namely, an individual constant for a particular time series, replenishes fractal analysis. The results of evaluations of this fractal constant for several time series are presented.

The questions of neural networks usage for cryptographic information protection have been considered. The problem of cryptographic stability and the main reasons for the unreliability of cryptosystems have been investigated. The issue of neurocomputers application in cryptanalysis is considered. The analysis of the neurocomputers work based on the developed software is done.

This article is dedicated to develop a model that correlates the relevant QoS parameters and QoE factors with impact on the variation of the user's perception of the quality. To achieve this goal, a process was planned and designed, taking into account all method and recommendations, duly described in the survey of the related work . New objective method of assessing the quality of the video based on the perception of the relationship detected taking into account the properties of self-similar traffic has been proposed.