Filtyp
With ongoing strains and disparities in the healthcare system, Generative AI applications are increasingly viewed as promising tools to alleviate pressure in this sector. Trust is seen as a critical enabler for the integration of GenAI tools. However, this area remains largely unexplored, with limited research examining user-perceived trust. This study addresses this gap by adopting a qualitative
This study investigates deep learning approaches for automated land use classification from high-resolution remote sensing imagery, comparing Convolutional Neural Network (CNN) and Vision Transformer (ViT) architectures. Eight semantic segmentation models were evaluated on the HRSCD dataset containing 291 aerial image pairs (0.5m resolution) with four land use classes: building, agricultural, fore
The transition towards renewable energy technologies is essential to mitigate climate change. However, the intermittency of Variable Renewable Energy (VRE) such as wind and solar creates seasonal imbalances in the electricity supply. To address this, inter-seasonal energy storage solutions are needed. Produced via electrolysis from excess electricity, hydrogen offers a promising storage medium. Ho
NMDA receptors are ligand-gated ion channels assembled as heterotetramers composed of glycine-binding GluN1 and GluN3 subunits (GluN3A-B), and glutamate-binding GluN2 subunits (GluN2A-D). Their involvement in various psychiatric and neurological disorders makes them potential therapeutic targets. However, compared to the extensively studied GluN1/2 NMDARs, the functional, physiological, and pharma
Recent advances in Large Language Models (LLMs) offer a promising alternative by enabling code generation from natural language. However, despite progress, LLMs still struggle with spatial reasoning, structural fidelity, and robustness across diverse GIS datasets. This thesis systematically compares three LLM adaptation strategies, i.e., prompt engineering, retrieval-augmented generation (RAG), an
This project introduces the development of a digital educational platform designed to strengthen knowledge and awareness of Parkinson’s disease among healthcare personnel, particularly within home care and nursing services. The initiative stems from a recognized need for accessible and disease-specific information, as new employees often do not receive the training and knowledge needed to provide
Historically, sleep stage classification has relied on manual scoring performed by experienced professionals. In recent years, the emergence of machine learning has enabled the automation of this process, thus saving both time and resources. Many of these sleep staging models depend on data which already has been annotated by a professional, but a few try to exploit the vast amount of unlabelled d
Human Activity Recognition (HAR) from wearable sensors is often constrained by limited labeled data and varying device placements. This thesis investigates a Transformer-based self-supervised approach that learns representations via a masked reconstruction and noise-injection pretext task, followed by fine-tuning on smaller labeled datasets. Experiments on WISDM, REALWORLD, and OPPORTUNITY confir
An optical system for digital light processing car headlamps was designed and optimized in this project. The specifications include five lens elements and a field of view of 15 degrees horizontally and 7.5 degrees vertically. The system consists of an illumination module and a projection lens assembly. A high-brightness LED source and a rectangular light integrator ensure uniform illumination on t
This Master's Thesis examines the possibility of developing a software tool for optimizing offshore wind power plant layouts in regards to wake effects. The study was conducted with the purpose of making development and operation of offshore wind power plants more efficient and cost effective. The tool, named OptiWind, was developed in Python and employs several imported packages for mathemati
S-Rprop is an extension of resilient backpropagation (Rprop), which introduces mini-batch learning to the Rprop algorithm. This is done by using two different mini-batch sizes, one for the learning rate and one for the weight updates, allowing for an accurate learning-rate update whilst gaining the fast weight updates of the stochastic gradient descent (SGD) optimizer. In this paper, the concept o
Emerging storage devices, particularly ferroelectric tunnel junction (FTJ) memristors, offer a compelling solution for high-density storage as they form a type of non-charge based non-volatile storage capable of storing multiple bits in each device; however, their adoption is hindered by reliability concerns. This work addresses this by presenting a design and implementation of an efficient Error
BastuBaћar is a public sauna experience carved into the limestone coast of Sliema, Malta - a reimagining of the long-lost Chalet, once a beloved landmark of leisure and gathering. This project introduces an unfamiliar, experimental ritual to the Maltese Islands - a guided sauna and bathing experience open to all. A synthesis of the Swedish word for sauna and the Maltese word for sea, the project
We explored various methods for inducing quantum dots (QDs) in pure zinc blende (ZB) InAs nanowires. This material exhibits high conductivity due to the presence of an intrinsic accumulation layer of charge carriers, resulting from surface band bending and donor impurities. To create regions that can be depleted by gate potentials (barriers) in the axial direction of the nanowire, we employed both
This thesis investigates the performance of two model predictive control (MPC) variants for autonomous racing, using a 1:10-scale vehicle as a testbed for reference path-tracking experiments. The platform is equipped with a single inertial navigation system (INS) sensor, providing highly accurate odometry measurements. The controllers under study are the standard gradient-based MPC and a sampling-
The aim of this thesis is to explore how methods from survival and event history analysis, namely relative risk regression, can be used to estimate prior probabilities in criminal cases, based on the distance between the potential perpetrator’s residence and the crime scene as well as further characteristics of the potential perpetrator. The locations of potential perpetrators residences can be mo
This study explores how young people's presence on social media can increase their risk of being recruited into criminal networks. It focuses on digital vulnerability, the normalization of crime, and challenges in preventive efforts. The research is based on semi-structured interviews with professionals from the police, social services, and civil society. It also includes an analysis of two ke
Denna uppsats undersöker den svenska lagändring som från och med juli 2023 innebär att alla som döms för narkotikaförsäljning av normalgraden ska få minst sex månaders fängelse. Reformen är en del av regeringens satsning på att motverka gängkriminalitet och öppen droghandel, och syftar enligt pro-position 2022/23:53 till att avskräcka såväl yrkeskriminella som unga rekryter från att delta i narkotThis thesis analyses the 2023 Swedish legislative reform that introduces a mandatory minimum prison sentence of six months for drug dealing, as stipu-lated in the Penal Law on Narcotics Offences (narkotikastrafflagen). The re-form forms part of a broader strategy against gang-related crime and open drug markets, with explicit governmental expectations of greater deterrence, disruption of criminal
This thesis explores advancements in Automatic Speaker Verification (ASV) by examining the impact of multiple speaker enrollments and introducing Adaptive Neural Probabilistic Linear Discriminant Analysis (Adaptive NPLDA). Modern ASV combines front-end feature extraction, using state-of-the-art methods based on Deep Neural Networks (DNNs), such as the ReDimNet architectures, with back-end modeling
