https://www.um.edu.mt/library/oar/handle/123456789/132533 Wed, 08 Apr 2026 11:54:36 GMT 2026-04-08T11:54:36Z https://www.um.edu.mt/library/oar/handle/123456789/145333 Title: Solving the inverse shortest path problem for earthquakes’ motion Abstract: According to Fermat’s principle, seismic waves follow paths of least travel time. Thus, shortest path algorithms such as Dijkstra’s can be used to determine these paths. Conversely, inferring the parameters of a mathematical program from an observed optimal path defines the inverse shortest path problem, an area within inverse optimisation. With its wide range of applicability, inverse optimisation has attracted considerable interest. One of the earliest topics in this field was the inverse shortest path problem, with Burton and Toint (1992) laying its foundations. Since then, this problem has been explored across several domains, with various mathematical formulations and algorithms proposed. This dissertation examines the inverse shortest path problem in depth, reviews its theoretical foundations, and applies it to a seismological case study. Three algorithms are employed to solve the problem: the column generation algorithm, a quadratic programming algorithm, and a deep inverse optimisation algorithm using a modern deep learning framework. The aim is to estimate the weight vector using these algorithms, thereby reconstructing the mathematical program that defines the shortest paths taken by seismic waves. To the best of the author’s knowledge, this is the first study to apply the inverse shortest path problem to local seismic data from the Maltese Islands and Sicily. Description: M.Sc.(Melit.) Wed, 01 Jan 2025 00:00:00 GMT https://www.um.edu.mt/library/oar/handle/123456789/145333 2025-01-01T00:00:00Z https://www.um.edu.mt/library/oar/handle/123456789/145332 Title: Seismic source characterization in the central Mediterranean Region Abstract: The process of earthquake source characterization involves determining the key properties of an earthquake and its origin, including its location, as indicated by latitude and longitude coordinates, its depth and its magnitude. This dissertation focuses on establishing a statistical model, by combining different Neural Network (NN) architectures including the Multi-Layer Perceptron (MLP), Convolutional Neural Network (CNN) and Graph Neural Network (GNN), in order to characterize the source of the earthquakes occurring in the central Mediterranean region, particularly concentrating on the Maltese islands, Sicily and the surrounding areas. The models considered are trained and validated on earthquake data recorded between 2013 and 2024. These data were obtained from seismic stations positioned around the Maltese Islands and Sicily, which are installed and maintained by the Istituto Nazionale di Geofisica e Vulcanologia (INGV) and the University of Malta’s Seismic Monitoring and Research Group (SMRG). The objective is to predict earthquake source parameters using station coordinates and waveform features. Each earthquake is represented as a graph in which the vertices correspond to stations, and the associated features are assigned to these nodes. For each event, the model outputs latitude, longitude, depth, and magnitude. Two model architectures are considered, namely, an edgeless graph model and a dynamic edges GNN. To identify the optimal model of each architecture, a systematic series of experiments, including hyperparameter tuning, regularization techniques, restricting the analysis to a more localized region and ensemble modelling, are conducted. The best two models are then fit on test data comprising events from January 2025 to August 2025. The edgeless graph architecture emerged as the best-performing architecture. Description: M.Sc.(Melit.) Wed, 01 Jan 2025 00:00:00 GMT https://www.um.edu.mt/library/oar/handle/123456789/145332 2025-01-01T00:00:00Z https://www.um.edu.mt/library/oar/handle/123456789/145327 Title: Traffic control optimization using Markov decision processes Abstract: Traditional traffic light systems often operate inefficiently, especially in high-intensity traffic situations, due to their reliance on a fixed cycle (FC). These traditional traffic light systems follow predetermined cycles, switching between red, yellow and green signals at fixed intervals, without accommodating real time traffic conditions. As a result of inefficient time allocation, traffic jams often build up. Throughout the years, researchers approached this problem using different techniques, many of which using the Markov Decision Process (MDP) framework. The MDP is a framework for analysing the existence and structure of good policies, which are self-thought rules behind choosing an action and for devising procedures for finding such policies. The aim of this dissertation is to identify policies that help manage traffic efficiently by considering various factors contributing to traffic congestion. This is achieved by analysing the evolution of the state of the intersection and by applying MDP based policies to have dynamic control of the traffic light system. The results in this study illustrate that an MDP-driven cycle significantly outperforms a traditional FC, particularly during high-intensity traffic conditions, when it comes to managing traffic efficiently. This is demonstrated in this dissertation by virtue of the MDP-model that has been developed and formulated originally by the author. Description: M.Sc.(Melit.) Wed, 01 Jan 2025 00:00:00 GMT https://www.um.edu.mt/library/oar/handle/123456789/145327 2025-01-01T00:00:00Z https://www.um.edu.mt/library/oar/handle/123456789/145326 Title: A study on human pose classification using convolutional neural networks and tensor regression Abstract: Folk dances are a source of a country’s history and traditions, and their documentation and analysis are important for their preservation. Human Pose Classification (HPC) covers the classification of human poses through body part detection from image, video, or measurement data. Folk dances are defined as a repeated sequence of main choreographic steps. The classification of the main choreographic steps of a dance falls under choreographic modeling, which is an application of HPC in dance. In this paper, we explore appropriate methods for choreographic modeling using image data, where we cover Convolutional Neural Networks (CNNs) and Tensor Regression (TR). CNNs are well-known in image classification since they are constructed more efficiently than Artificial Neural Networks (ANNs) for working with image data. TR is an extension of the regression problem using tensor representations, which would be more appropriate than classical regression for use with image data. We do a comparison study between CNNs and TR on a dance dataset, aiming to predict all poses over two trials. The first trial uses a standard training-test split across all dancers, while the second follows a leave-one-out approach, training on all but one dancer and testing on the excluded dancer. Both models correctly predicted all poses in the first trial, whereas the second trial proved more complicated, with fewer poses classified. CNNs yielded higher performance metrics compared to TR, where TR generally had worse results. However, CNNs contained significantly more parameters than TR, leading to signs of overfitting in the second trial. Description: M.Sc.(Melit.) Wed, 01 Jan 2025 00:00:00 GMT https://www.um.edu.mt/library/oar/handle/123456789/145326 2025-01-01T00:00:00Z