Högskolebiblioteket i Halmstad Publikationer
Not a member yet
17134 research outputs found
Sort by
Prototype Aligned Embedding for Time Series Forecasting
Time-series forecasting plays a crucial role in supporting data-drivendecisions across domains like energy systems, finance, and healthcare.Despite their strong predictive power, many deep learning models arestill viewed as "black-boxes", making it difficult to understand howthey arrive at their predictions. This lack of transparency often limits their broader adoption in real-world applications. To address thischallenge, this thesis proposes an interpretable, prototype-based deeplearning framework for multivariate time-series forecasting that integrates both numerical and textual modalities to enhance predictiveaccuracy and transparency. At the core of the proposed framework lies the Prototype Similarity Network (ProSeNet), which learns a set of representative temporal patterns called prototypes that serve as interpretable anchors formodel reasoning. Each forecast is explained in terms of similarity tothese learned prototypes, providing clear insights into the model’sdecision process. To further enhance contextual understanding, theframework is extended into a multimodal architecture where textualembeddings derived from pretrained language models are fused withtemporal features, enabling the model to reason jointly over structured and unstructured data. Comprehensive experiments demonstrate that the proposed approachachieves strong predictive performance while maintaining interpretability. The prototype-based design allows users to trace predictions backto meaningful temporal patterns, and the integration of textual embeddings significantly improves contextual reasoning and accuracy.This work contributes a transparent, context-aware forecasting framework that bridges interpretable representation learning and multimodal reasoning advancing the development of trustworthy and explainable AI systems for time-dependent applications
Exploring real-time sensor fusion of 3D LiDAR and RGB-D streams on NVIDIA Jetson AGX Orin
This thesis investigates real-time sensor fusion of 3D LiDAR and RGBD data streams on embedded platforms, with a focus on adaptive gated networks. While most of the state-of-the-art 3D object detection models are designed for high-end GPUs, embedded devices such as the NVIDIA Jetson AGX Orin are constrained by power, space, and computational resources. To address this, we implemented and evaluated multiple fusion strategies, early, mid, late, and hybrid approaches. The models were deployed on the Jetson platform and tested in real-world indoor environments using a mobile robotic platform. The experimental setup deployed pre-trained detection heads, namely TR3D and VoteNet, and a custom dataset consisting of chairs, tables, and bookshelves. The fusion strategies were benchmarked in terms of mean Average Precision (mAP) at various Intersection over Union (IoU) thresholds, alongside inference time (FPS) as a measure of real-time feasibility. Results show that early fusion improves accuracy compared to camera-only baselines while reducing inference time by compressing data before processing. Late fusion achieves the highest detection accuracy across all classes but at the cost of significantly reduced inference speed, as it requires running multiple models in parallel. Mid-fusion alone underperformed in detection accuracy compared to the baselines but contributed positively in hybrid configurations, particularly when combined with late fusion, yielding the best balance between accuracy and efficiency. The findings indicate that the embedded device cannot yet achieve full real-time performance of 10 FPS without further optimization, such as TensorRT support for sparse tensors. However, the fastest fusion model on the Jetson device approaches real-time performance at 8.7 FPS, although its accuracy is 77.4% lower (0.091 mAP) compared to the best-performing model, which achieves 0.403 mAP at 4.07 FPS. This work contributes to a comparison of fusion strategies on real-world data and provides insights into the trade-offs required for deploying multi-modal perception in resource-constrained robotic systems
Manipulatives in Mathematics Education : A Systematic Literature Review on the Impact of Manipulatives on Student’s learning
Laborativt material utgör en viktig del av matematikundervisningen. Många elever upplever utmaningar och svårigheter med matematik i olika termer och former och tar därför hjälp av laborativt material. Syftet med denna studie var att ta reda på hur laborativa material påverkar elevers lärande, med fokus på elevers prestation, begreppsförståelse och motivation. Flera studier har undersökt både konkret material och dess påverkan på elever samt lärare, men även på hur virtuellt laborativt material påverkar eleverna och lärarna. Vi valde därför att undersöka båda materialen från både lärarnas perspektiv samt elevernas perspektiv för att få en inblick i hur materialet påverkar eleverna. Genom en systematisk litteraturstudie identifierades åtta relevanta artiklar som användes i studien. Resultaten analyserades tematiskt och delades upp i två huvudteman med sex kategorier. Studien visar att både konkreta och virtuella laborativa material kan bidra till förbättrad prestation och ökad förståelse av matematiska begrepp. Konkret material framstår som särskilt gynnsamt för att belysa samband mellan begrepp och att upptäcka och bearbeta missförstånd. Virtuella material ger andra didaktiska möjligheter som variation, dynamiska representationer och snabb återkoppling, vilket hjälper elevernas progression mot mer abstrakt tänkande. Resultatet visade också att materialen bidrar till ökat engagemang och motivation hos elever.Manipulatives play an important role in mathematics education. Many students experience challenges and difficulties in mathematics in various ways and therefore benefit from the use of manipulative materials. The aim of this study was to investigate how manipulatives affect students’ learning, with a focus on academic performance, conceptual understanding, and motivation. Several studies have examined the use of concrete manipulatives and their impact on both students and teachers, as well as how virtual manipulatives influence learning and teaching. Therefore, this study examines both types of materials from both teachers’ and students’ perspectives in order to gain insight into how these materials affect students’ learning. Through a systematic literature review, eight relevant articles were identified and included in the analysis. The results were analysed thematically and organised into two main themes with six categories. The findings indicate that both concrete and virtual manipulatives can contribute to improved performance and increased understanding of mathematical concepts. Concrete manipulatives appear to be particularly beneficial in highlighting relationships between concepts and in identifying and addressing misconceptions. Virtual manipulatives offer additional didactical opportunities, such as variation, dynamic representations, and immediate feedback, which support students’ progression towards more abstract thinking. The results also show that the use of manipulatives contributes to increased student engagement and motivation
The positions of Swedish parliamentary parties regarding gender-discriminatory advertising
Denna studie analyserar hur svenska riksdagspartier konstruerar problemet könsdiskriminerande reklam och hur dessa problemformuleringar legitimerar olika politiska lösningar. Studien utgår från Carol Bacchis policyanalytiska ansats What’s the Problem Represented to Be? (WPR), som undersöker hur könsdiskriminerande reklam definieras, vilka antaganden som tas för givna samt vilka konsekvenser dessa problemrepresentationer får. Studien kombinerar WPR-analysen med ett ideologiskt teoretiskt ramverk baserat på feministisk, liberal och konservativ ideologi för att kategorisera hur ideologiska utgångspunkter formar partiernas syn på reklam, ansvar och reglering. Analysen baseras på riksdagsmotioner och offentliga politiska uttalanden från samtliga svenska riksdagspartier. Resultaten visar att partier med en feministisk ideologisk orientering tenderar att konstruera könsdiskriminerande reklam som ett strukturellt samhällsproblem som reproducerar ojämna könsmaktsrelationer och därför kräver statlig reglering. Partier med mer liberala och konservativa orienteringar framställer istället problemet som begränsat, individuellt eller möjligt att hantera genom marknadsbaserad självreglering, vilket legitimerar en fortsatt tilltro till Reklamombudsmannen. Konservativa partier argumenterar dock för att det befintliga självreglerande systemet bör kompletteras med lagstiftning. Studien visar att skillnader i partiernas ställningstaganden i hög grad kan förklaras av ideologiska utgångspunkter snarare än av oenighet kring huruvida könsdiskriminerande reklam utgör ett samhällsproblem.This study analyses how Swedish parliamentary parties construct the problem of gender-discriminatory advertising and how these legitimize different political solutions. This study is based on Carol Bacchi’s policy-analytical approach What’s the Problem Represented to Be? (WPR) that examines how gender-discriminatory advertising is defined, which assumptions are taken for granted and what consequences these problem formulations produce. The study combines WPR analysis with an ideological theory framework based on feminist, liberal and conservative ideologies in order to categorize how ideological standpoints shape parties views on advertising, responsibility and regulation. The analys is based on parliamentary motions and public political statements from all Swedish parliamentary parties. The results show that parties with feminist ideological orientations tend to construct gender-discriminatory advertising as a structural societal problem that reproduces unequal gender power relations and therefore requires state regulation. Parties with more liberal and conservative orientations instead frame the problem as limited, individual or manageable through market-based self-regulation, thereby legitimizing continued reliance on the advertising “Reklamombudsmannen”. However, conservative parties argue that the existing self-regulatory system should be modified through a legal complement. The study demonstrates that differences in party positions are largely rooted in ideological standpoints rather than in disagreement over the existence of gender-discriminatory advertising as a societal issue.
Outdoor mathematics pedagogy : A literature study on pupils’ learning, motivation and pedagogical preconditions
Trots utökad undervisningstid i matematik visade både nationella och internationella mätningar att svenska elevers matematikresultat stått still eller sjunkit, särskilt för elever med matematiska svårigheter. Detta väckte frågor om vilka didaktiska metoder som användes i matematikundervisningen samt hur de kunde organiseras och struktureras. Syftet med vår litteraturstudie var att undersöka hur utomhuspedagogik påverkat mellanstadieelevers matematikutveckling och dess pedagogiska förutsättningar. Litteraturstudien byggde på en tematisk genomgång av sakkunniggranskade artiklar publicerade mellan åren 2017–2024, identifierade på databasen ERIC (EBSCO) och på söktjänsten OneSearch. Resultatet visade att utomhuspedagogik i matematik främst användes som ett didaktiskt komplement till katederundervisning, samt att utomhuspedagogik bidrog positivt till bearbetningen av den abstrakta och begreppsliga förståelsen i matematik. Det framkom även hur utomhuspedagogisk undervisning bidrog till ökad motivation, engagemang och fokus hos elever. Utomhusundervisningen stärkte samarbetet mellan elever och relationer mellan elev och lärare. Studien pekade på att elevers resultat till stor del var beroende av lärarens utbildning, erfarenhet, förmåga att bygga struktur samt länka samman inomhus- och utomhusundervisning. Slutligen indikerade resultatet att utomhuspedagogisk matematikundervisning hade potential att stärka matematikutvecklingen och pedagogiska förutsättningar för mellanstadieelever.Despite increased instructional time in mathematics, both national and international assessments show that Swedish students' mathematics achievement remains stagnant or is declining, particularly among students experiencing mathematical difficulties. This raises questions about which didactic approaches are used and how mathematics instruction is organized and structured. The aim of this literature review is to examine how outdoor education influences lower elementary school students' mathematical development and the pedagogical conditions required for its successful implementation. The study is based on a thematic review of peer-reviewed articles published between 2017 and 2024, identified through the database ERIC (EBSCO) and on the search service OneSearch. The results show that outdoor education in mathematics is primarily used as a didactic complement to its traditional classroom-based instruction and that it contributes positively to the development of abstract and conceptual understanding. The findings also indicate increased student motivation, engagement, and focus, as well as strengthened collaboration between peers and improved teacher-student relationships. Furthermore, the study demonstrates that students’ learning outcomes largely depend on teachers’ education, experience, and ability to create structure and to coherently link indoor and outdoor instruction. Overall, the results suggest that outdoor mathematicseducation has the potential to enhance mathematical development and pedagogical conditions for lower elementary school students
Governing AI : A study of problem representation and governance logics in the EU: AI - Regulation
The purpose of this study is to examine how the EU AI Act (2024) has been adopted as a regulatory framework, with the principal focus being on how “problems” are represented. The thesis examines how the AI Act represents and addresses specific problems associated with AI, such as disinformation and deepfakes, biometric mass surveillance, discrimination and bias, privacy risks, cybersecurity, and robustness of the digital infrastructure. The thesis also examines which risks are accentuated, which are marginalized and “unspoken”, and how knowledge and power shape the governing rationality of the European Union regarding AI regulation. The analysis is conducted through the application of Carol Bacchi’s What’s the Problem Represented to be? (WPR) method in combination with a Governmentality-perspective. Through the use of qualitative document analysis, the European AI Act is analyzed to identify dominant discourses, underlying problem assumptions, and the effects produced by these. Our findings indicate that the regulation mainly represents AI-related risks as technical and manageable through “risk-based” classification, compliance mechanisms and certain transparency obligations. Although certain issues, such as discrimination and surveillance are recognized, they are primarily addressed through technical precautions rather than social, political and economic interventions at a structural level. Other risks, such as power disparities within the AI industry and broader democratic implications such as public- and institutional trust - remain largely under-problematized and “forgotten”. Finally, the thesis concludes that the EU AI Act serves, not only as a judicial instrument, but also as a tool for governance, which shapes how risks regarding AI are understood, represented and addressed. Through a combination of risk rationalities and habitual values, the European Union also accomplishes a certain indirect governance by defining acceptable practices, responsibilities and legitimate knowledge surrounding AI.
The Arms Industry and Arms Exports in Swedish Security Policy after NATO Accession: How Are the Arguments Reframed? : An Argumentation Analysis of State Actors and the Swedish Peace and Arbitration Society (Svenska Freds) after Sweden’s Nato Accession
Denna studie undersöker och analyserar hur den officiella argumentationen kring den svenska vapenindustrin och vapenexporten har förändrats i samband med Sveriges historiska omläggning av säkerhetspolitiken och anslutningen till Nato. Syftet med uppsatsen är att bidra till en fördjupad förståelse av hur statliga aktörer legitimerar och rättfärdigar vapenexport i en ny säkerhetspolitisk kontext, samt hur dessa nya narrativ utmanas och ifrågasätts av det civila samhället. Studien genomförs metodologiskt som en argumentationsanalys av ett strategiskt urval statliga dokument, däribland utrikesdeklarationer och skrivelser om strategisk exportkontroll, från tidsperioderna 2021-2022 samt 2024-2025. Detta ställs i relation till kritiska publikationer och debattinlägg från Svenska Freds- och Skiljedomsföreningen. Det teoretiska ramverket baseras på Johan Galtungs teoretiska ansats om positiv och negativ fred, vilket används för att belysa och analysera de motstridiga säkerhetslogiker som präglar debatten. Resultaten visar att argumentationen har genomgått ett fundamentalt och genomgripande skifte. Före ansökningsprocessen till Nato påbörjades, motiverades behovet av en inhemsk vapenindustri primärt genom argument om nationellt oberoende och militär alliansfrihet. Efter inträdet i Nato legitimeras verksamheten istället genom hänvisningar till allianssolidaritet, interoperabilitet och kollektiv avskräckning. Analysen indiker att staten i allt högre grad prioriterar en logik baserad på “negativ fred” (militär styrka och frånvaro av våld) vilket står i direkt konflikt med de krav på “positiv fred” (demokrati, mänskliga rättigheter och hållbar fred) som förespråkas av Svenska Freds- och Skiljedomsföreningen. Studiens slutsats är att Nato-medlemskapet har medfört att realpolitiska säkerhetsintressen har fått företräde framför normativa värderingar, vilket resulterade i en omdefiniering av Sveriges internationella roll. Från att vara en neutral fredsnation till att bli en lojal säkerhetsproducent inom alliansen. This study examines how the official argumentation regarding the Swedish arms industry and arms exports has changed in connection with Sweden’s accession to Nato. The purpose is to contribute to a deepened understanding of how state actors legitimize arms exports in a new security policy context, and how these arguments are challenged by civil society. The study is conducted as a qualitative argumentation analysis of state documents (Statements of Foreign Policy and communications on strategic export control) from the periods 2021-2022 and 2024-2025, as well as critical publications from the “Svenska Freds- och Skiljedomsföreningen”. The theoretical framework is based on Johan Galtung’s concepts of positive and negative peace to illuminate conflicting security logics. The results show that the argumentation has undergone a fundamental shift. Prior to the accession, the arms industry was motivated by the need for national independence and military non-alignment. Following the entry into Nato, it is instead legitimized through concepts of alliance solidarity and collective deterrence. The analysis indicates that the state prioritizes a logic of “negative peace” (military strength), which conflicts with the demand for "positive peace” (democracy and human rights) advocated by the “Svenska Freds- och Skiljedomsföreningen”. The conclusion is that Nato membership has entailed that “realpolitik” security interests have taken precedence over normative values, resulting in a redefinition of Sweden’s role from a neutral peace state to a loyal security provider
A Deepfake Detection with CNN-Based Feature Extraction Method
This research develops a deep learning model for detecting Deepfakes by leveraging modern convolutional neural networks (CNNs) for feature extraction and a Support Vector Machine (SVM) for classification. The model achieves powerful performance metrics, shows that it has the ability to protect the integrity of digital content and fight false information. The goal is to assist in locating and eliminating inaccurate information in media content. The key advantages of this approach include the elimination of parameter tuning, efficient low scale feature selection, and the suitability of SVM for binary classification problems, offering better performance with fewer computational resources. By combining CNNs for feature extraction and SVM for classification, this method provides a dependable, computationally efficient solution for deepfake detection. The model is evaluated using accuracy, F1 score, and the confusion matrix to ensure robust performance in distinguishing real from fake content
Text-based discussions' impact on reading comprehension : A literature review on the effects of text-based discussions about fiction on students' reading comprehension
PIRLS senaste undersökning visade att elevers läsförståelse har försämrats. Detta gav upphov till vår fråga om hur lärare kan arbeta för att förbättra elevers resultat. Syftet med denna studie var att undersöka hur textsamtal om skönlitteratur påverkar elevers läsförståelse. Studien genomfördes som en systematisk litteraturstudie där åtta vetenskapliga artiklar valdes ut och analyserades. Genom en tematisk analys av dessa artiklar identifierades tre huvudteman: lärarens roll, textsamtal och litteraturundervisning. Förutsatt att textsamtalen var välstrukturerade samt att de tillhörande frågorna var väl genomtänkta framhölls textsamtal som en gynnsam undervisningsmetod. PIRLS' latest study showed that students' reading comprehension has deteriorated. This prompted our question on how teachers can work to improve students' results. The purpose of this study was to investigate how literature-based discussions around fictional texts affect students' reading comprehension. The study was conducted as a systematic literature review, where eight scientific articles were selected and analysed. Through a thematic analysis of these articles, three main themes were identified: the teacher's role, literature-based discussions, and literature teaching. Provided that the discussions were well-structured, and the accompanying questions carefully considered, literature-based discussions were highlighted as a beneficial teaching method
The collaboration within the domains of the Swedish language classroom and school library : A literature study which investigates whether the collaboration between school librarians and teachers promotes pupils' reading development
I den svenska samhällsdebatten lyfts det ofta att Sverige, med sjunkande resultat i nationella och internationella undersökningar, befinner sig i en läskris. Flera politiska förslag har lagts fram och en del av dessa har genererat förändringar i svensk skola. Däribland en förändring i skollagen som syftar till att stärka skolbiblioteken och elevernas tillgång till dessa. Syftet med denna studie var att undersöka hur läraren och skolbibliotekarien kan främja elevers läsutveckling genom att verka i varandras domäner samt vilken betydelse detta har när svenskläraren utformar sin undervisning. Till grund för studien ligger sju vetenskapliga artiklar samt två doktorsavhandlingar som har analyserats och sedan tematiserats. Studiens resultat visade att ett skolbibliotek i sig inte är tillräckligt för elevers läsutveckling. För att läsutvecklingen ska främjas behöver svenskläraren och skolbibliotekarien bedriva ett framgångsrikt samarbete. Den påverkan som skolbibliotekarien har på svenskundervisningen är också beroende av ett framgångsrikt samarbete. Det räcker inte med att den enskildes profession existerar, de behöver samexistera gemensamt i svensklärarens didaktik. Forskningen visar att organisering och ledning av samarbetet utgör viktiga förutsättningar för att främja elevers läsutveckling. Samtidigt menar forskningen att det finns en bristande vilja till att samarbeta samt en otillräcklig förståelse för varandras professioner vilket kan utgöra hinder för ett gemensamt arbete. Within the public debate in Sweden, it is often highlighted that the results of both national and international surveys, related to the Swedish pupils’ reading ability, are drastically declining. Due to pupils’ poor literacy, this has led to what the politicians define as a national reading crisis. As a result of this, changes in the Swedish school system are taking place, among them a change in the law which aims to strengthen the school libraries and make them accessible to all pupils. The purpose of this study was to investigate how the collaboration between the Swedish language teacher and the school librarian promotes pupils’ reading development. Furthermore, the study identifies the significance of these collaborations within the Swedish teachers' didactics. The study is based on seven scientific articles and two doctoral theses , which have been analysed and thematized. Previous research of the collaboration shows that organization and leadership are important essentials for improving students' reading development. The research also presents that there is a lack of willingness to collaborate and a limited understanding of each other's professions, which can act as barriers for collaboration. The findings indicate that a school library, on its own, does not significantly enhance students' reading development. In order to improve reading development, Swedish teachers and school librarians must collaborate successfully. Moreover, the impact that school librarians have on Swedish language teaching also depends on successful collaboration. It is not enough for the individual professions to exist, they need to coexist within Swedish teachers' didactics