Veranstaltungsprogramm

<p>The integration of deep learning, federated learning, and Large Language Models (LLMs) has led to the development of an innovative framework known as Federated learning-based Large Language Model (Fed_LLM). This framework optimizes learning systems and enhances privacy protections by leveraging advanced computational techniques to achieve precise academic guidance and robust data privacy safeguards. The Fed_LLM method combines these technologies to provide tailored educational solutions while striving to maintain high levels of privacy protection. Deep learning algorithms are employed to analyze complex and immense educational datasets, for extracting accessible insights and features from the data (Li et al., 2020). Federated learning introduces a decentralized model training approach that significantly enhances data confidentiality. It keeps sensitive user or learner information on local devices, minimizing privacy risks and ensuring that personal data remains protected (Yang et al., 2019; Carlini et al., 2020). Simultaneously, LLMs improve the system's capacity to understand and process natural language inputs, enhancing user interaction and communication.</p>

<p>This study addresses the integration of deep learning, federated learning, and Large Language Models (LLMs) to optimize learning systems while ensuring data privacy. The Fed_LLM framework uses Google's Federated Averaging (FedAvg) for privacy-preserving model updates and Transformer-based LLMs for processing complex language data. This approach enables personalized learning in diverse university settings, enhancing efficiency and engagement without compromising privacy (Kairouz et al., 2021; Li et al., 2020; Carlini et al., 2020; Yang et al., 2019). To evaluate the effectiveness, efficiency, and security of Fed_LLM, a detailed set of performance metrics is required. These include traditional machine learning metrics like Accuracy, Error Rates, F1 Score, Precision, and Recall, along with language-specific metrics such as Perplexity and BLEU Score for evaluating large language models. Additionally, metrics specific to federated learning such as communication efficiency and the number of communication rounds are used to ensure privacy. Computational efficiency metrics like training time and inference latency are also used (Carlini et al., 2020; Yang et al., 2019). This methodology promises to provide a step towards a secure and efficient learning environment that meets the needs of future generations of learners. By combining advanced technologies to protect user privacy while providing personalized learning experiences, Fed_LLM offers a promising future.</p>

The global shift towards digitalization has profoundly impacted various industries, with the sports industry being no exception. Businesses have witnessed a significant transformation driven by the integration of modern technologies and the reshaping of traditional business operations. This shift, commonly referred to as digital transformation (DT), necessitates rapid adaptation efforts to keep up with the evolving digital landscape of consumer demands, market dynamics, and technological advancements. Existing research has revealed significant gaps in comprehensively addressing the multifaceted nature of undergoing DT within the sports industry. Therefore, this study aims to address this gap by examining the digital capabilities crucial for driving DT, with a specific focus on German sports entities. Through a multi-perspective analysis involving semi-structured expert interviews the investigation explores how digital capabilities can be leveraged and implemented to undergo digital change. Qualitative content analysis was conducted on the insights and best practices provided by real-life industry representatives from professional sports clubs and organizations to uncover patterns and derive novel insights contributing to understanding how sports entities can thrive in the digital world. The findings reveal that the extent to which digital capabilities can be implemented varies due to differences in available resources, financial capacities, and organizational readiness. However, by assessing their individual circumstances and understanding the obstacles to transformation, clubs and organizations can effectively harness digital capabilities tailored to their specific organizational context to ultimately foster innovation and guide sports entities towards sustainable success and competitiveness in the dynamic digital landscape.

<p>Motivation: In modern working environments, which are characterized by the principles of New Work and globalization, the importance of effective group dynamics is increasing. Isolated group members can impair the functionality and success of teams. Understanding the emergence and effects of isolation in group phases according to Tuckman (1965) and identifying solutions are therefore of crucial importance.</p>

<p>Goal: This study aims to identify the different types of isolation in online collaborative environments and to analyze how data-driven methods can contribute to the detection and integration of isolated group members. The findings from scientific papers are brought together to provide a comprehensive overview of existing solutions. Methods: A systematic literature review (SLR) was conducted using the PRISMA model to capture relevant studies from 2015 to 2024 from the databases Web of Science and EEE. The analysis focused on behavioral patterns and algorithms aimed at detecting isolated members in online groups. Discussion: The study highlights the need to implement data-driven approaches to recognize and support isolated members in online groups. These approaches can not only enhance group performance, but also improve individual well-being and promote social cohesion. Limitations: The study encounters limitations of systematic literature review such as publication bias and variability in data quality, which may influence the results. In addition, the adaptation of the identified behavioral patterns to different online collaboration environments is seen as a necessary future task. </p>

<p>Implications: The implications of this study are significant for the design of modern work environments. They enable organizations to develop accurate tools that provide real-time feedback on group interactions, allowing isolation to be quickly identified and preventative action to be taken. This not only improves team performance by actively engaging all members, but also promotes the working environment and social cohesion. It also strengthens the inclusive work culture, especially in globally diverse teams, and contributes to scientific research by revealing specific behavioral patterns and algorithms for identifying isolated group members. These findings open up new ways of optimizing teamwork and overcoming the challenges of the digital world of work. <br /></p>