Portal del coneixement obert de la UPC

Enviaments recents

• Carregant...

  Projecte Final de Màster Oficial Accés obert

  A game theory approach to set thresholds for influence spread

  (Universitat Politècnica de Catalunya, 2026-01-28) García Fernández, Jaya; Universitat Politècnica de Catalunya. Departament de Ciències de la Computació

  Més...

  Social networks are the natural space for the spreading of information and influence and have become a media themselves. Several models capturing that diffusion process have been proposed, most of them based on the Independent Cascade (IC) model or on the Linear Threshold (LT) model. Although the LT-based models contemplate an individual threshold for each actor in the network, the existing experimental studies so far have almost always considered identical thresholds for all the actors. Our main objective in this work is to initiate the study on how the dissemination of information on networks behaves when those thresholds are set strategically by the actors. For doing so, we plan to design different non-cooperative games incorporating characteristics of the game, and analyze problems related to the expansion of influence in their Nash equilibrium network or by the use of dynamics on the best response graph.

  Més...
• Carregant...

  Projecte Final de Màster Oficial Accés obert

  Using LLMs for real-time multimedia-based cybersecurity analysis

  (Universitat Politècnica de Catalunya, 2026-01-28) Kosinski, Bartosz; Universitat Politècnica de Catalunya. Departament de Ciències de la Computació

  Més...

  Nowadays, large language models (LLMs) are widely used across nearly all domains, and data indicate that their adoption continues to grow each year. A fascinating aspect of artificial intelligence in today's world lies in its contextual capabilities, not only to search for information, as traditional browsers do, but also to utilize that information to generate meaningful and context-aware outputs based on given inputs. However, LLMs also present certain disadvantages that are particularly critical in domains such as cybersecurity. Most notably being the reliability and accuracy of generated information, which can be compromised by hallucinations leading to misinformation and flawed decision-making in processes such as incident classification or mitigation. This master's thesis focuses on the application of LLMs in the field of cybersecurity, with particular emphasis on their use by analysts, SOC teams, and other security professionals. The research addresses the need for a locally deployable, privacy-preserving, and explainable LLM framework capable of assisting professionals with text and document based tasks in a safe and effective manner. While cloud-based deployments remain common, they raise persistent concerns regarding privacy, data control, and the reliability of model outputs in high-security environments. Moreover, the thesis evaluates the applicability of the proposed system by identifying and assessing potential cybersecurity tasks for which it can be effectively employed through a comparative evaluation of outputs generated by different models integrated into the proposed prototype.

  Més...
• Carregant...

  Projecte Final de Màster Oficial Accés obert

  Gammification and user study of a virtual reality application for assisted rehabilitation

  (Universitat Politècnica de Catalunya, 2026-01-28) Gregori Barrera, Victor; Universitat Politècnica de Catalunya. Departament de Ciències de la Computació

  Més...

  This thesis explores the use of Virtual Reality (VR) in the physiotherapeutic field, focusing on the rehabilitation of patients with frozen shoulder. Traditional rehabilitation is often perceived as tedious and painful, which can negatively effect the motivation and adherence of the patient to the treatment. This thesis investigates how using a virtual world with gamified exercises may improve patient engagement during rehabilitation exercises. In previous studies, it has been shown that Virtual Reality applications can help in patient performance and motivation. Additionally , some experts in the field consider it useful and find it interesting to use them for actual patient rehabilitation. This thesis improves an already existing VR application implemented to track hand position and assess patient progress. Two different additional modes were implemented: a gamified version of the original exercises and a "placebo" version that emulates the traditional exercises typically performed in the doctor consultation. A pilot user study was conducted to evaluate user behavior, opinions, and perceived engagement across both modes. The results provide insightful potential, showing that the gamified VR exercises increased perceived user engagement, immersion and enjoyment compared to the placebo version, while maintaining exercise clarity and correctness. These findings suggest that gamified VR can be a promising tool to enhance motivation and adherence in frozen shoulder rehabilitation, paving the way for future clinical studies.

  Més...
• Carregant...

  Projecte Final de Màster Oficial Accés obert

  Predictive mapping of indoor radon in Spain using graph neural networks

  (Universitat Politècnica de Catalunya, 2026-01-28) Loobuyck, Senne; Universitat Politècnica de Catalunya. Departament d'Arquitectura de Computadors

  Més...

  Indoor radon exposure represents a major environmental health risk and is a leading cause of lung cancer among non-smokers. Because direct radon measurements are time-intensive and costly, radon risk assessment in Spain largely relies on national radon potential maps that characterise geogenic hazard but do not account for building-specific factors that strongly influence actual indoor radon concentrations. In this dissertation, graph neural networks (GNNs) are investigated as a data-driven framework for predicting indoor radon levels in Spain by jointly modelling geogenic context and building-level characteristics. The proposed approach integrates heterogeneous information sources, including geogenic information such as radon potential and lithology, together with building attributes such as floor level, construction material, building age, and season of measurement. Spatial relationships between observations are represented at the postal-code level, enabling explicit modelling of spatial dependencies that are ignored by traditional tabular methods. Several graph-based representations are proposed, ranging from aggregated postal-level graphs to building-level homogeneous and heterogeneous architectures, and are evaluated on binary and multiclass radon classification tasks. In this setting, the study is structured around two complementary evaluation objectives. First, different graph model variants are assessed under spatial cross-validation and region-wise hold-out experiments in order to identify architectures that generalise well across regions. The results indicate that, given the limited dataset size, simpler homogeneous graph models consistently outperform more complex heterogeneous variants, underscoring the importance of aligning model complexity with data availability. Second, the selected homogeneous graph model is compared with strong non-graph tabular baselines, including Random Forests (RF), Support Vector Machines (SVM), and the national radon potential (RP) map. In the binary classification setting, the GNN consistently outperforms RP, although the improvement remains modest, while the trainable tabular baselines do not substantially improve upon the RP map. In contrast, in the multiclass setting, the performance of RP deteriorates sharply, whereas all trainable models (both graph-based and tabular) achieve markedly superior results. This highlights the limitations of geogenic-only indicators for fine-grained indoor radon categorisation. Overall, this work demonstrates that graph neural networks offer a principled and effective framework for indoor radon prediction when spatial dependencies and building-level factors are central to the task, improving upon geogenic-only indicators such as radon potential in settings where additional contextual information is available. At the same time, it underscores that graph-based models are not universally superior and must be evaluated carefully against simpler alternatives.

  Més...
• Carregant...

  Projecte Final de Màster Oficial Accés obert

  Network-aware Federated learning

  (Universitat Politècnica de Catalunya, 2026-01-28) Chen, Xi; Universitat Politècnica de Catalunya. Departament d'Arquitectura de Computadors

  Més...

  Federated learning (FL) enables collaborative model training across distributed clients while preserving data privacy, but it suffers from substantial communication overhead and sensitivity to heterogeneous client data and participation. In this thesis, we systematically investigate the federated learning process from both server-side orchestration and client-side training perspectives, with particular emphasis on how client participation patterns and data heterogeneity influence convergence behavior and communication efficiency. To address the high communication cost inherent in FL, we propose FedDB, a distance-based client selection strategy. FedDB evaluates the relative contribution of individual clients using model distance metrics computed at the server and selectively regulates client participation across training rounds. Importantly, the proposed method operates entirely at the server side and requires no modification to client-side training procedures or model architectures. We evaluate FedDB across multiple learning scenarios, including convolutional neural network (CNN) training and large language model (LLM) fine-tuning. Experimental results demonstrate that FedDB achieves significant communication savings-reducing client participation frequency by up to 60% compared to baseline federated learning strategies-while maintaining competitive predictive performance and stable convergence. Overall, this work highlights the effectiveness of dynamic, distance-based client selection for communication-efficient federated learning in distributed system.

  Més...