Mykolo Romerio universitetas / Mykolas Romeris University
Deep sequence learning with multi-task supervision for scalable population health monitoringItem type:Publication, research article[2026][S1a][N009][21] ;Ashraf, Imran; ;Awais, Muhammad ;Bouchelligua, Wided ;Mansour, Sahar ;Alyounis, EssaNawaz, MajedFrontiers in public health, 2026, vol. 14, p. 1-21Models designed for artificial intelligence-driven population health monitoring must be able to integrate multiple types of information over time, be heterogeneous in the data they consider, and be deployed at scale. This research proposes an integrated multitask framework for predicting population disease incidence and mortality risk, based on deep learning and using both longitudinal biobank and National Health Survey data in a scalable manner. The framework will learn shared temporal representations from clinical lab-generated/demographically defined survey variables by applying a strict prospective evaluation approach across the framework. On UK Biobank, the proposed model achieves an AUROC of 0.842 and a C-index of 0.791, outperforming strong baselines such as Proformer having AUROC 0.829 and UKB-MDRMF having AUROC 0.821. On NHANES, the model achieves an AUROC of 0.861 and a C-index of 0.808, exceeding those of commonly used boosted tree and survival-based approaches. Temporal robustness analysis shows controlled performance degradation across extended horizons, with UK Biobank AUROC decreasing from 0.872 at 1 year to 0.819 at 10 years and NHANES AUROC decreasing from 0.886 at 1 year to 0.836 at 5 years. Subgroup evaluation demonstrates stable performance across age, sex, and clinical risk strata, while scalability analysis confirms near-linear training growth and stable inference latency below three milliseconds per individual at biobank scale.
1 Hackathons as a Method for Fostering Corporate Innovation and Competitive AdvantageItem type:Publication, research article[2025][P1a2][S003,N009][9]; ; ; The Proceedings of the 20th European Conference on Innovation and Entrepreneurship, hosted by University of the National Education Commission, Krakow, Poland, 25-26 September 2025., p. 322-330Hackathons, originally rooted in the software and technology sectors, have evolved into powerful tools for corporate innovation. Today, organizations across diverse industries use them to accelerate idea generation, boost employee engagement, and strengthen competitive positioning. However, despite their growing popularity, the mechanisms by which hackathons drive innovation within firms remain underexplored in academic literature, particularly from the standpoint of innovation theory and strategic management. Much of the existing research is fragmented across disciplines such as healthcare, education, and computer science, lacking a unified framework for understanding hackathons as instruments of organizational innovation. This study addresses that gap by examining hackathons as structured innovation development mechanisms. Drawing on a comprehensive review of scholarly sources, we investigate the definitions, theoretical foundations, and practical implementations of corporate hackathons. The paper develops a conceptual model that delineates the phases, design components, critical success factors, and post-event integration processes necessary to foster sustainable innovation outcomes. Our findings suggest that hackathons support corporate innovation by enabling rapid experimentation, attracting entrepreneurial talent, and cultivating internal knowledge-sharing communities. Moreover, they provide a low-risk environment conducive to intrapreneurship and openness to change. This article contributes to the academic discourse on innovation and organizational learning by situating hackathon practices within strategic innovation frameworks. It also offers actionable insights for practitioners on how to effectively design, implement, and leverage hackathons to drive competitive advantage in dynamic and uncertain business environments.
16 19Scopus© Citations 1 Contactless Fall Detection Using mmWave Radar with Stacking-Based Ensemble LearningItem type:Publication, research article[2025][S5][N009,T007,S003][9] ;Lee, Young Sil; ; Jung, Sang-JoongThe Journal of Korea Institute of Convergence Signal Processing., 2025, p. 315-323With the transition to an aging society, increasing attention has been directed toward early detection of falls among elderly individuals. Vision-based fall detection systems suffer from privacy concerns and environmental sensitivity, while wearable sensor-based approaches impose a user burden. To address these limitations, this paper proposes a contactless fall detection method using millimeter-wave (mmWave) radar. Radar data collected in an indoor environment are used to design a stacking-based ensemble learning framework that combines traditional machine learning and deep learning models. The proposed method integrates predictions from multiple base classifiers through a meta-classifier to improve detection performance and robustness. Experimental results using real-world radar data demonstrate that the proposed ensemble model outperforms conventional single classifiers across accuracy, precision, recall, and F1-score. These findings indicate that radar-based sensing, combined with ensemble learning, provides a practical, privacy-preserving solution for indoor fall detection.
22 LLM-Based Learning for Enhancing Interaction Between Game Players and NPCsItem type:Publication, research article[2025][S5][N009,T007,S003][9] ;Huofeng, Li; ;Zhou, JianiYu, TaesooThe Journal of Korea Institute of Convergence Signal Processing., 2025, p. 191-199In today's games, traditional non-player character (NPC) behaviors and dialogues rely on preset scripts, making it difficult to dynamically respond to diverse player needs and limiting the immersion and interactive depth of the game. This study aims to explore the application of Large Language Models(LLM) in games to enhance the intelligent interaction of NPCs. In this study, the Llama model is fine-tuned to build a simulated game scene in Unity engine, and the fine-tuned model is embedded into an NPC system to verify its performance and ability to drive NPCs in real game interactions. In the experiments, the research team designed a series of typical game interaction scenarios, including item collection, task delivery, and complex semantic Q&A, through which the NPC's language comprehension and interaction capabilities were tested. The results demonstrate that LLM-driven NPCs enable game developers to build more flexible and intelligent interaction mechanisms. This approach opens new possibilities for intelligent NPC development.
2 Public Governance of Sustainable Development Implementation: A Case Study of Coastal Municipalities in LithuaniaItem type:Publication, research article[2025][S1][N009,S004,S003][26]; ;Kiyak, Deimena ;Olšauskas, VirgilijusSustainability, 2025, p. 1-26This study aims to answer the research question what are the characteristics of existing institutional governance factors within Lithuanian coastal municipalities that lead to the establishment of a common aim of sustainable development across different coastal municipalities. The method employed includes a systematic academic literature review and document analysis. The data collected were coded according to predefined themes and sub-themes, facilitating the determination of extent to which institutional factors influencing the public governance of sustainable development are integrated into the structures and credentials of municipalities. The findings reveal that while some municipalities incorporate sustainability into their vision and strategic plans, there is a general lack of dedicated policies and accountability for sustainable development. The absence of a dedicated unit or responsible employee for sustainable development implementation further fragments the approach to addressing these issues. Communication methods and public engagement vary, impacting the effectiveness of sustainable development initiatives. This study concludes that meta-governance serves as a unifying factor in public governance coordination among municipalities. To foster sustainable development in Lithuanian coastal cities, recommendations for policymakers are provided to increase transparency and civic trust, lead initiatives, and enhance dialogue through regional stakeholder forums or digital platforms for the co-creation of solutions across coastal municipalities.
6 16Scopus© Citations 2 A Predictive Model for Contemporary Art Investment: Integrating Geopolitical Risk Factors through Artificial Intelligence and Machine Learning TechniquesItem type:Publication, conference paper[2024][T2][N009,S003][1]; ;Palikaridis, Tilemachos ;Karampidis, KonstantinosPapadourakis, GiorgosAmi-Es 2024. International Symposium "Ambient Intelligence and Embedded Systems". 11-14 September, 2024, Tallin university of Technology (TallTech), Tallin, Estonia, p. 1The global art market has become increasingly complex and volatile, particularly in regions affected by geopolitical conflicts. This research aims to develop a novel approach to art investment strategies by leveraging artificial intelligence and machine learning techniques. The study focuses on contemporary art from geopolitically sensitive regions, including Ukraine, Russia, Israel, Palestine, Azerbaijan, Armenia, and Georgia. Utilizing a comprehensive contemporary art prices database (Artsy), we employ advanced AI methods to extract relevant data and apply machine learning algorithms to analyze market trends and generate predictive models. Our primary objective is to investigate the extent to which an artist's nationality and other factors associated with conflict zones influence art price fluctuations in the global market. The research seeks to create a robust decision-making model to assist potential investors in evaluating artwork as a financial asset. By incorporating geopolitical considerations into the investment analysis, we aim to provide a more nuanced understanding of risk and potential returns in the contemporary art market. This interdisciplinary project combines computer science, management, and art history expertise to address the growing need for sophisticated investment tools in an increasingly uncertain global landscape. The findings of this study have the potential to inform individual and institutional investment strategies and contribute to the broader discourse on the intersection of art, management, and geopolitics.
14 The future-proof arteficial intelligence (AI) and the resilient digital and energy infrastructure nexusItem type:Publication, conference paper[2024][P2a][N009,T001][4]Науково-практична конференція «Резильєнтність динамічних систем» : Програма конференції, 27 грудня 2024 року, Kyiv., p. 1-4About 8000 data centers are worldwide (33% - the USA, 16% - the EU, 10% -China). Between 2017 and 2021 (EPRI, 2024), electricity used by Meta, Amazon, Microsoft, and Google for data center cloud computing doubled. In the USA, data centers will consume up to 9.1% of electricity annually by 2030 versus about 4% in 2024. The total ICT energy consumption in the world can rise from 2-3% up to 16% for the next decades. The growth of data centers and the adoption of artificial intelligence (AI) rely on the availability of electric power. Opportunities for investors in power infrastructure and adjacent sectors are quickly arisen. Electricity operating expenditures comprise about 20 percent of the total cost base for data center business models, which have proved highly profitable for large companies. The US government (gov) currently classifies AI as a national security issue (House, 2024). It also addresses additional sensitive security issues, including countering adversary use of AI. The EU targeted 75% of European enterprises to leverage cloud computing, big data, and AI to drive innovation and efficiency by 2030 (The European Commission, 2023). In 2024, the UA gov and the Ministry of Digital Transformation issued the concept (КОНЦЕПЦІЯ Державної цільової науково-технічної програми з використання технологій штучного інтелекту в пріоритетних галузях економіки на період до 2026 року, 2024), the road map (Ministry of Digital Transformation in Ukraine, 2023), and the white paper (Ministry of Digital Transformation in Ukraine, 2024) about the AI future. The special focus is on the state AI program for 2024-26, including legislation. The World Bank survey has shown that the ICT-energy nodes represent one of the key risks during the war.
6 Prediction of the Main Environmental and Energy Characteristics of Diesel Engines using an Artificial Neural Network for Pure Diesel Fuel, Pure HVO, and a Mixture of these Fuels in the Ratio 50/50 by VolumeItem type:Publication, research article[2024][S1][N009,T003,T001][14] ;Matijošius, Jonas ;Rimkus, AlfredasTransport problems = Problemy transportu., 2024, p. 69-82This research assesses the influence of different quantities of hydrotreated vegetable oil (HVO) in diesel fuel on the performance of the engine and the emissions it produces. The particular areas of interest are the level of smoke emitted and the brake thermal efficiency (BTE). A series of engine experiments were undertaken to quantify emissions and performance characteristics under various operating regimes using three distinct fuel blends: D100 (pure diesel), HVO50 (50% HVO mixture), and HVO100 (pure HVO). The results show a tendency to reduce soot emissions when the amount of HVO in the mixture reaches 50%, but in order to accurately determine the correlation between the amount of HVO and emissions, additional studies with various concentrations of HVO and diesel mixtures are necessary. Similarly, the results show a slight improvement in BTE stability with a 50% HVO blend, but more studies with different percentages of HVOdiesel blends are needed to reliably determine changes in BTE. This indicates a trend of change in engine performance with increasing HVO concentration in diesel fuel. In order to forecast emissions and performance indicators under different operating scenarios, we used artificial neural networks ANNs, which demonstrated excellent prediction accuracy, exhibiting robust linear correlations between the expected and real values for all fuel types. This research emphasizes the advantages of utilizing HVO in diesel engines for both environmental impact and performance. It also emphasizes the usefulness of ANNs in optimizing engine settings to improve efficiency and minimize emissions. The results endorse the further use of HVO as a viable substitute for conventional diesel, leading to less ecological consequences and enhanced engine efficiency.
9Scopus© Citations 3 - research article[2024][S1][T003,T001,N009][37]
;Matijošius, Jonas ;Rimkus, AlfredasMachines: Special issue: Cutting-edge technologies and applications in automatic control systems., 2024, p. 1-37.Machine learning models have been used to precisely forecast emissions from diesel engines, specifically examining the impact of various fuel types (HVO10, HVO 30, HVO40, HVO50) on the accuracy of emission forecasts. The research has revealed that models with different numbers of perceptrons had greater initial error rates, which subsequently reached a stable state after further training. Additionally, the research has revealed that augmenting the proportion of Hydrogenated Vegetable Oil (HVO) resulted in the enhanced precision of emission predictions. The use of visual data representations, such as histograms and scatter plots, yielded significant insights into the model’s versatility across different fuel types. The discovery of these results is vital for enhancing engine performance and fulfilling environmental regulations. This study highlights the capacity of machine learning in monitoring the environment and controlling engines and proposes further investigation into enhancing models and making real-time predictive adjustments. The novelty of the research is based on the determination of the input interface (a sufficient amount of input parameters, including chemical as well as technical), which characterizes the different regimes of the diesel engine. The novelty of the methodology is based on the selection of a suitable ANN type and architecture, which allows us to predict the required parameters for a wide range of input intervals (different types of mixtures consisting of HVO and pure diesel, different loads, different RPMs, etc.).
11 25Scopus© Citations 9 Modeling of anxiety spread during the pandemicsItem type:Publication, conference paper[2022][T2][S006,N009][1] ;Denisov, Vitalij; Sakalauskas, LeonidasModelling and Simulation of Social-Behavioural Phenomena in Creative Societies (MSBC-2022) : 2nd International & European Conference, September 21-23, 2022 Vilnius, Lithuania : the book of abstracts. Hamburg : Hamburg University of Applied Sciences, 2022., p. 14The study aims to develop a hybrid model of pandemic anxiety and panic dynamics, calibrated through indirect social anxiety and emotion level indicators, in order to apply it to the analysis, forecasting and management of anxiety and related panic scenarios. The developed model combines agent modeling, dynamic systems modeling with differential equations and machine learning methods. Because direct investigation of the prevalence and spread of anxiety and related panic is a complex task, coronavirus morbidity data and indirect anxiety indicator data will be used to calibrate and verify the developed model, which will be processed by machine learning algorithms. One of the possible reflections of public anxiety is the web content. The sentiment analysis will identify and extract subjective information about anxiety and emotional level from the web media content. In this way, the application of machine learning methods would help to identify additional variables and their properties in the development of a hybrid model.
13