Repositório Científico de Acesso Aberto da ULisboa
Repositório Institucional da Universidade de Lisboa
Entradas recentes
Utilizing GNNs to predict 3D convex polyhedra from 2D planes
Publication . Schmitz,Niklas; Faculty of Sciences; Department of Informatics; Araújo,Nuno Miguel Azevedo Machado de; Pesquita,Cátia Luísa Santana Calisto
This thesis tackles the folding problem: starting from a 2D template of rigid, hinge-linked faces, pre-dict which face pairs should be adjacent in the final structure, purely topologically, so that the tem-plate closes into a convex 3D polyhedron. The main challenge is the large combinatorial search space, which grows quadratically with face count, making rule-based approaches difficult to scale. This motivates investigating data-driven methods, specifically Graph Neural Networks (GNNs),to assess their suitability for folding. The work constructs a data set pairing polyhedral with their unique 2D unfoldings, spanning 2039 convex polyhedral and about 1.6M unfoldings. Building on this, I present FINE-GNN (Fixed-nodeI Ncremental EdgeGNN), an adapted GNN architecture that per-forms sequential, node-centric edge predictions. At each step, it decides whether and where to add an edge, providing a general architecture for edge prediction on fixed node sets that extends beyond folding. Using this architecture, I inject degree-budget constraints derived from the face’s shape. This simple topological signal cuts false positives, lifts Precision, and enables exact-match recon-structions at lower face counts (predicted adjacency equals ground truth), something prior GNNs fail to achieve. I further study node-ordering policies, which determine the next face to process. While macro metrics (Precision, Recall) differ little, per-graph analyses reveal clear distinctions: Structured policies consistently outperform a random order. The strongest model configuration in this thesis, using constraints with a cyclic order policy, achieves 37.2% Precision overall, with much higher val-ues at lower face counts (78.3% at 6 faces, 61.2% at 7, 47.0% at 8).These results show that GNNs are a promising fit for folding at small to moderate sizes and offer a practical foundation for scaling to larger, more complex polyhedra. To my knowledge, this is the first data-driven study to directly predict face adjacencies for polyhedral folding, establishing a base line for subsequent research.
Building resilient landscapes in the Mediterranean: knowledge for agriculture and forestry. Putting knowledge into action: practical solutions and demonstrations
Publication . Paulo,Joana Amaral; Firmino,Paulo Neves; Department of Natural Resources, Environment and Territory; Forest Research Centre
Modelos de volatilidade: Aplicação a dados das bolsas de valores de Lisboa e de Madrid
Publication . Fernandes,Ana Filipa Rodrigues; Faculdade de Ciências; Sotelo,Luis Gimeno; Bermudez,Patrícia Cortés de Zea
Time series analysis enables a clearer understanding of how a variable evolves over time, allowing for the identification of behavioural patterns and the use of this information to forecast future values. In the financial sector, the importance of analysing time series is even greater, as it allows investors to assess risk and estimate returns based on more informed and strategic decisions. However, the high volatility (conditional variance) typical of these series makes the evolution of future prices unstable. Therefore, studying volatility is essential to measure the risk associated with financial assets. The aim of this work is to apply ARCH and GARCH volatility models to the time series of the PSI and IBEX 35 indices between 2011 and 2024. The study seeks to determine which method is most efficient for modelling and forecasting the volatility of the Portuguese and Spanish markets and to compare their results. The ARCH(1) model is generally insufficient to capture conditional heteroskedasticity, as it requires a high number of lags, leading to a non-parsimonious model. Thus, the GARCH(1,1) model is the most appropriate to describe the variance of the processes, since it can model more persistent volatility by incorporating both past shocks and past variance. When the GARCH model adequately captures the variance but fails to account for autocorrelation in the mean, a joint ARMA–GARCH estimation is recommended. The results showed that the best model for the PSI, in both modelling and forecasting, was the ARMA(2,2)–GARCH(1,1) with a Student’s t-distribution, due to the strong autocorrelation of log-returns and heavy tails. For the IBEX 35, the GARCH(1,1) model with a Student’s t-distribution proved adequate, indicating no dependency in the mean. It was observed that the forecast volatility of both indices rises initially and then stabilises at the long-term value, displaying a somewhat unrealistic pattern due to the model’s difficulty in capturing asymmetries and shocks.
BUGSPOTTING 2.0 – Criação de datasets de vulnerabilidades com recurso a inteligência artificial
Publication . Gonçalves,David Miguel Dias; Faculdade de Ciências; Departamento de Informática ; Neves,Nuno Fuentecilla Maia Ferreira; Medeiros,Ibéria Vitória de Sousa
Technological advancements and increasingly faster information processing challenges the Effective functioning and security of software programs and web applications. Int his context, web applications are the primary targets chosen by attackers, aiming to exploit potential vulnerabilities they may contain. It has become a key and essential objective to prevent, detect, and mitigate these vulnerabilities quickly and effectively. Static code analysis tools (Static Analysis Tools, SASTs) have been widely used for vulnerability detection through the inspection of application source code. These tools perform a fast execution analysis with low maintenance costs, as maintaining and modifying the code is easier. However, they generate a high number of false positives and negatives, making it necessary to implement additional methods alongside their analysis. Although SASTs are used by highly knowledgeable professionals, they are not perfect and are prone to errors, as previously mentioned. Recent studies have shown that the use of Machine Learning (ML) techniques can assist and enhance the development of these tools, improving their effectiveness [8]. In this regard, to apply these techniques, it is necessary to build a precise and reliable dataset to train ML models. This work is based on the BugSpotting1.0 website, which allows for the classification of slices (pieces of PHP code) both by SASTs and through crowdsourcing. In this dissertation, we propose a new approach to building datasets, as well as improving the classify cation algorithms of slices, which will enable the creation of reliable datasets and the use of ML models for discovering vulnerabilities in web applications written in PHP. These new implementations result in a new version of BugSpotting, version 2.0.
Les français sous l’uniforme allemand pendant la Seconde Guerre Mondiale 1940-1945
Publication . Cathelin,Louis; Universidade de Lisboa; Faculdade de Letras; Ventura,António Adriano de Ascenção Pires; Varandas,Jose Manuel Henriques
Esta dissertação tem por objetivo principal estudar as motivações dos soldados franceses, saídos de um contexto e de uma guerra que os opôs à Alemanha do III Reich, para integrar o exército alemão enquanto voluntários. Uma apresentação analítica dos contextos militar e político entre alemães e franceses daquele período é outro dos nossos propósitos, desde a criação da Grande Alemanha até ao início da Segunda Guerra Mundial. Um terceiro momento da nossa investigação procura incidir sobre a perspetiva com que foi vista a derrota militar francesa e o momento da instauração de um novo governo saído do armistício e propício às ideias alemãs. Outro ponto de focagem desta dissertação será a análise e observação das mudanças no Exército francês e nas demais instituições francesas, sobretudo num plano de análise da colaboração militar e económica Franco-Alemã.O estabelecimento de um objetivo comum contra um inimigo político e militar, representado pelas forças aliadas, permitirá a introdução de uma primeira força de combate franco-alemã chamada "Legião dos Voluntários Franceses" subordinada ao exército regular alemão, e cuja formação estudaremos. Outro ponto de observação tem a ver com a incorporação e os primeiros compromissos militares, enquanto força de combate na Frente Leste, que não satisfarão as autoridades alemãs e colocaram esta Legião num plano de menor importância, com as consequências inerentes a esse estado. Um último aspeto a desenvolver nesta dissertação será o da integração daquela Legião sob as bandeiras da Waffen-SS, até à sua dissolução, e reformulação numa unidade que lutou independentemente pela sua sobrevivência num Reich moribundo. Esta foi a divisão Charlemagne, que combateu em circunstâncias extremas até à vitória dos aliados em maio de 1945.