https://www.um.edu.mt/library/oar/handle/123456789/135418 Sat, 18 Jul 2026 16:07:36 GMT 2026-07-18T16:07:36Z https://www.um.edu.mt/library/oar/handle/123456789/146931 Title: Leveraging invariant prediction for mitigating specificity constraints in affect modelling Abstract: Affect modelling aims to predict human emotional states from multimodal signals, yet current approaches often struggle to generalise beyond the specific datasets or contexts in which they are trained. This dissertation investigates the use of in variant features, predictors whose relationship with affective states remains stable across distinct environments, as a strategy to improve generalisability. To this end, two publicly available corpora, AGAIN and RECOLA, were systematically parti tioned into environments defined by user, task, and annotator triplets. An envi ronment refers to the conditions under which data is collected, and data gathered within the same environment is assumed to come from the same underlying distri bution. The Invariant Causal Prediction (ICP) framework was employed to identify stable features across these environments, which were then compared against full feature sets and principal components derived through PCA. Three supervised learning models—Logistic Regression, a feed-forward Neural Network,and a Long Short-Term Memory (LSTM)network — were trained under all three feature conditions, using group-based cross-validation to avoid information leakage. Results demonstrate that invariant features can deliver measurable benefits for feed-forward models, particularly in enhancing accuracy and correlation while substantially reducing feature dimensionality. However, their advantages were less consistent for sequence models like LSTMs, where temporal dependencies were not fully captured by invariants alone. Statistical significance tests further showed that invariant features improved balanced classification (F1) more strongly in AGAIN than in RECOLA,underscoring the dataset-specific nature of their effectiveness. Overall, the findings highlight both the promise and the limitations of invariance in affect modelling. While not a universal solution, invariant features represent a principled means of isolating robust predictors across heterogeneous contexts, contributing to the broader goal of developing affective systems that are reliable, interpretable, and adaptable across diverse real-world settings. Description: M.Sc.(Melit.) Wed, 01 Jan 2025 00:00:00 GMT https://www.um.edu.mt/library/oar/handle/123456789/146931 2025-01-01T00:00:00Z https://www.um.edu.mt/library/oar/handle/123456789/146930 Title: Siamese network‐based vector embeddings of MRI scans for twin identification Abstract: Monozygotic twins are identical twins that develop from a single fertilised egg that spontaneously splits, resulting in two individuals sharing 100% genetic material. Identifying monozygotic twins from brain MRI scans represents a frontier challenge in computational medical imaging with significant implications for understanding genetic influences on neuroanatomical structure through direct pattern recognition. While classical twin studies using ACE models decompose statistical variance to establish independent regional heritability estimates (60‐80%), this study introduces a fundamentally different computational framework that learns directly from MRI data to rank neuroanatomical regions by their collective discriminative capacity for genetic similarity detection, complementing traditional statistical approaches through data‐driven analysis. Adeep learning methodology employing Siamese networks with 3D CNN backbones is developed for automated twin identification using 138 genetically verified monozygotic twin pairs (276 subjects) from the Human Connectome Project S1200 dataset. Modified U‐Net, ResNet, and DenseNet architectures generate 128‐dimensional embeddings optimised via triplet loss with hard negative mining, forcing models to learn subtle genetic signatures by focusing on challenging discriminative examples that distinguish twins from their most similar morphological matches. U‐Net achieved superior computational performance with 92.0% F1‐score (σ = 2.5%), 95.2% AUC‐ROC, and 91.4% accuracy, while ResNet demonstrated competitive results (89.6% F1‐score) and DenseNet showed greater variability (88.5% F1‐score). Embedding analysis reveals clear bimodal separation between genetically related and unrelated individuals through learned morphological patterns. Layer‐Wise Relevance Propagation analysis provides the first data‐driven ranking of neuroanatomical regions by discriminative importance for genetic relatedness detection. Statistical analysis reveals pronounced subcortical dominance with large effect size (Cohen’s d = 2.80, p = 3.89e‐6), with six subcortical structures occupying top positions, including the thalamus (0.955), brainstem (0.875), and hypothalamus (0.707). This computational hierarchy contrasts with traditional ACE studies reporting highest heritability in cortical areas (frontal 78‐95%, temporal 77‐89%), demonstrating that direct pattern recognition from MRI data identifies different neuroanatomical signatures than statistical variance decomposition. Notably, models utilise practically all brain regions (most importance scores > 0.2), indicating distributed multivariate processing rather than selective regional dependence. Ablation studies confirm data augmentation’s critical role, with substantial i performance improvements across CNN architectures. Clinical integration through standard neuroimaging formats in Connectome Workbench demonstrates immediate practical utility, positioning this computational approach for adoption in research and clinical environments requiring direct analysis of genetic influences in brain structure. The framework advances precision neuroimaging by providing automated, quantitative genetic similarity detection through direct pattern recognition, revealing spatial insights that complement traditional heritability studies while offering methodological advances applicable to diverse medical imaging classification tasks requiring regional discriminative analysis Description: M.Sc.(Melit.) Wed, 01 Jan 2025 00:00:00 GMT https://www.um.edu.mt/library/oar/handle/123456789/146930 2025-01-01T00:00:00Z https://www.um.edu.mt/library/oar/handle/123456789/146599 Title: Spell checking for the Maltese language Abstract: This study presents the development of a Grammar Error Correction (GEC) system for the Maltese language. A GEC system, or spell checking system, improves writing quality by identifying and correcting spelling and grammar errors in text. Modernspell checkers are able to improve writing across various contexts, ranging from casual text messages to formal documents. As a low‐resourced and under represented language in the digital world, Maltese lacks a robust digital presence, highlighting the urgent need for a dedicated spell‐checking system. This research seekstocontributetothedevelopmentofaspellcheckerfortheMalteselanguage. A key issue identified through previous efforts for GEC systems for Maltese is the lack of data available. Therefore, the creation of a larger, more representa tive dataset was necessary. A data collection campaign was launched to gather authentic human errors. The errors collected were statistically analysed, and used to inform the creation of a synthetic dataset. As a result, two distinct datasets— containing authentic human errors, synthetic errors, and a hybrid of both—were developed and used to train the system. The created system consisted of a trans former basedimplementation, inwhichpre‐trained Malteselanguagemodelswere implementedforboththeencoderanddecodercomponents. Thefinalsystemout performed previous spell‐checking systems, setting a new benchmark in Maltese GEC. The final system created consistently corrected errors related to capitalisation andMaltese‐specificcharacters, indicatingastronglevelofcontextualunderstand ing. Despite these advancements, thesystem’s overall performance remains below that of widely used commercial spell checkers. Nonetheless, the resources created and the findings of this study provide a foundation for future research in Maltese GEC. Description: M.Sc.(Melit.) Wed, 01 Jan 2025 00:00:00 GMT https://www.um.edu.mt/library/oar/handle/123456789/146599 2025-01-01T00:00:00Z https://www.um.edu.mt/library/oar/handle/123456789/146069 Title: Feedback loops and bias in machine learning algorithms for predictive policing Abstract: Predictive policing describes several emerging practices of implementing artificial intelligence and machine learning in police work, specifically in attempting to predict future crimes through algorithmic crime forecasting. These emerging practices have introduced many new opportunities for improved police work, but critics of predictive policing have raised both ethical and practical concerns. These concerns include the risk of feedback loops and bias. This thesis aims to contribute to this ongoing debate by examining how algorithmic crime forecasting tools produce bias and feedback loops and by exploring if it is possible to create algorithmic crime forecasting tools with reduced tendencies towards bias and feedback loops. Specifically, the focus is on the seminal and widely adopted PredPol system, which is based on an earthquake prediction system known as Epidemic Type Aftershock Sequence (ETAS). The methodology used in this studywas to replicate studies detailing the PredPol system, as well as studies criticising it. Based on previous findings by critics, a synthetic population and urn modelling was used to demonstrate the negative tendencies of the system. Based on this, an original framework was developed for evaluating modifications made to the algorithm by measuring the effectiveness in reducing feedback loop tendencies and improving fairness. This is done through metrics like the Predictive Accuracy Index (PAI), variations in the mean conditional intensity rates, λ, and the total fairness score, which evaluates the consistency of law enforcement attention across different demographic groups. To reduce the algorithm’s tendencies towards bias and feedback loops, a modified algorithm using rejection sampling and a fairness penalty was developed. While the proposed algorithmic adjustments lead to increased fairness and reduced feedback loop generation in predictive policing, they also introduce some trade-offs in predictive performance, particularly noted in the PAI values. However, the enhancements significantly mitigate biased policing practices and reduce the perpetuation of historical inequities, aligning more closely with ethical standards. Description: M.Sc. ICT(Melit.) Wed, 01 Jan 2025 00:00:00 GMT https://www.um.edu.mt/library/oar/handle/123456789/146069 2025-01-01T00:00:00Z