Publikationer från Örebro universitet
Not a member yet
18319 research outputs found
Sort by
KEA: Keeping Exploration Alive by Proactively Coordinating Exploration Strategies
No full textSoft Actor-Critic (SAC) has achieved notable success in continuous control tasks but struggles in sparse reward settings, where infrequent rewards make efficient exploration challenging. While novelty-based exploration methods address this issue by encouraging the agent to explore novel states, they are not trivial to apply to SAC. In particular, managing the interaction between novelty-based exploration and SAC’s stochastic policy can lead to inefficient exploration and redundant sample collection. In this paper, we propose KEA (Keeping Exploration Alive) which tackles the inefficiencies in balancing exploration strategies when combining SAC with novelty-based exploration. KEA integrates a novelty-augmented SAC with a standard SAC agent, proactively coordinated via a switching mechanism. This coordination allows the agent to maintain stochasticity in high-novelty regions, enhancing exploration efficiency and reducing repeated sample collection. We first analyze this potential issue in a 2D navigation task, and then evaluate KEA on the DeepSea hard-exploration benchmark as well as sparse reward control tasks from the DeepMind Control Suite. Compared to state-of-the-art novelty-based exploration baselines, our experiments show that KEA significantly improves learning efficiency and robustness in sparse reward setups.This work has received funding from the EU’s Horizon 2020 research and innovation programme under grant agreement No 101017274, and was supported by the Wallenberg AI, Autonomous Systems and Software Program (WASP) funded by the Knut and Alice Wallenberg Foundation.DARK
Combining Code Generating Large Language Models and Self-Play to Iteratively Refine Strategies in Games
No full textWe propose a self-play approach to generating strategies for playing in multi-player games, where strategies are represented as computer code. We use large language models (LLMs) to generate pieces of code to play in the game, which we refer to as generated bots. We engage the LLM generated bots in competitions, designed to generate increasingly stronger strategies. We follow game theoretic principles in organizing these tournaments, and use a Policy Space Response Oracle (PSRO) approach. We start with an initial set of LLM generated bots, and continue in rounds for adding new bots into the population. Each round adds a bot to the population by asking the LLM to produce code for playing against a bot representing the Nash equilibrium mixture over the current population. Our analysis shows that even a few rounds are sufficient to produces strong bots for playing the game. Our demo shows the process for the game of Checkers. We allow users to select initial bots in the population, run the process, inspect how the bots evolve over time, and play against the generated bots.Demo Track</p
Exploration of health-related quality of life, anxiety, and depression in antineutrophil cytoplasmic antibody-associated vasculitis
No full textOBJECTIVES: To explore health-related quality of life (HRQoL), anxiety, depression and fatigue among persons with ANCA-associated vasculitis (AAV). METHODS: In this cross-sectional study, patients were assessed with the EuroQoL five-dimension three-level questionnaire (EQ-5D-3L) and visual analogue scale (EQ-VAS), the Hospital Anxiety and Depression Score (HADS), and the Multidimensional Assessment of Fatigue questionnaire (MAF). RESULTS: HRQoL measured with EQ-VAS was higher in younger patients, those with inactive disease (BVAS = 0), and those with long disease duration (≥3 years), while EQ-5D-3L only varied with disease activity. Women, patients with active disease, and patients with shorter disease duration reported more anxiety, and younger persons with active disease and short disease duration reported more fatigue. In the total group, disease activity and disease duration were both associated with HRQoL, anxiety, depression and fatigue. Four clusters based on EQ-5D, EQ-5D index, and HADS were identified, containing patients with various levels of HRQoL and psychological distress across the disease course. CONCLUSION: In patients with AAV, HRQoL, anxiety, depression, and fatigue, is persistent but varies depending on disease activity, disease duration, gender, and age. Associations between disease activity and duration and HRQOL, anxiety, depression, and fatigue were present in all patients. Four clusters revealed the ongoing influence of AAV, emphasizing the continuous need for multiprofessional support during the disease course. In this study, EQ-VAS was better able than EQ-5D-index to effectively distinguish subgroups with different levels of HRQoL.Funding Agencies:Funding: This work was supported by the Swedish Rheumatism Association [ST-202111 to A.G.] [R-1025817 to I.G]. Stockholm County Council Regional Council (ALF) [FoUI –985152 to I.G]. The KingGustaf V 80-year Foundation [FAI 2023-0958 to I.G].</p
DFW : a novel weighting scheme for covariate balancing and treatment effect estimation
No full textEstimating causal effects from observational data is challenging due to selection bias, which leads to imbalanced covariate distributions across treatment groups. Propensity score-based weighting methods are widely used to address this issue by reweighting samples to simulate a randomized controlled trial (RCT). However, the effectiveness of these methods heavily depends on the observed data and the accuracy of the propensity score estimator. For example, inverse propensity weighting (IPW) assigns weights based on the inverse of the propensity score, which can lead to instable weights when propensity scores have high variance-either due to data or model misspecification-ultimately degrading the ability of handling selection bias and treatment effect estimation. To overcome these limitations, we propose Deconfounding Factor Weighting (DFW), a novel propensity score-based approach that leverages the deconfounding factor-to construct stable and effective sample weights. DFW prioritizes less confounded samples while mitigating the influence of highly confounded ones, producing a pseudopopulation that better approximates a RCT. Our approach ensures bounded weights, lower variance, and improved covariate balance.While DFW is formulated for binary treatments, it naturally extends to multi-treatment settings, as the deconfounding factor is computed based on the estimated probability of the treatment actually received by each sample. Through extensive experiments on real-world benchmark and synthetic datasets, we demonstrate that DFW outperforms existing methods, including IPW and CBPS, in both covariate balancing and treatment effect estimation
Infant Gaze Following Is Stable Across Markedly Different Cultures and Resilient to Family Adversities Associated With War and Climate Change
No full textGaze following in infancy allows triadic social interactions and a comprehension of other individuals and their surroundings. Despite its importance for early development, its ontology is debated, with theories suggesting that gaze following is either a universal core capacity or an experience-dependent learned behavior. A critical test of these theories among 809 nine-month-olds from Africa (Uganda and Zimbabwe), Europe (Sweden), and Asia (Bhutan) demonstrated that infants follow gaze to a similar degree regardless of environmental factors such as culture, maternal well-being (postpartum depression, well-being), or traumatic family events (related to war and/or climate change). These findings suggest that gaze following may be a universal, experience-expectant process that is resilient to adversity and similar across a wide range of human experiences-a core foundation for social development
Disability representation in the ITV crime drama Midsomer Murders : A balance between audience acceptance and alienation
No full textThis study explores disability representation in the transnational television crime drama Midsomer Murders(1997-present). It focuses on Lana who has a partial right arm and how she is positioned as a central figure in the Christmas 2023 episode. The analysis reveals a complex and contradictory pattern of interwoven ableist and anti-ableist ideas. Even though negative stereotypes are avoided, Lana’s portrayal is simultaneously visually and emotionally conditioned to avoid alienating audiences. The study speaks to critical scholars across disciplines with a broad interest in the fictional television crime drama and its abilities to communicate social (diversity) issues
Legal Rhetoric in Scandinavia : An Introduction and a Research Overview
No full textThis article provides an introduction to the field of legal rhetoric and an overview of Scandinavian research in the area. Legal rhetoric is an interdisciplinary field situated between law and rhetoric, concerned with how language is used to persuade in legal contexts. The article begins by tracing the shared historical roots of law and rhetoric, with a focus on ancient Greece and Rome, where rhetorical skill was integral to both legal and political life. It then identifies several key points of contact between the two disciplines today, including their shared emphasis on persuasion, contextual awareness, and linguistic craftsmanship. The article proceeds to present three main branches of contemporary legal rhetoric: the communicative, which focuses on practical rhetorical tools for lawyers; the methodological, which explores how rhetorical methods can support legal reasoning and analysis; and the philosophical, which investigates foundational questions at the intersection of rhetoric and legal theory. The third section offers an overview of Scandinavian contributions to the field, including anthologies, journal articles, doctoral dissertations, and textbooks. Particular attention is given to works that apply rhetorical theory - such as the theory of topics - to legal argumentation. The section highlights both pioneering publications and more recent developments in Denmark, Norway, and Sweden. In its concluding section, the article reflects on the current state and future of legal rhetoric in Scandinavia. It notes a growing interest among legal professionals and educators, as evidenced by the increasing presence of rhetoric in law school curricula and continuing legal education. At the same time, the article identifies areas - such as legal plain language - where rhetorical perspectives remain underutilized. The article concludes that legal rhetoric in Scandinavia has significant untapped potential, both as a research field and as a resource for legal practice.This text is a revised version of the article "Rättsretorik'', previously written for the Scandinavian research encyclopedia "Nordisk retorikkleksikon".</p
Possibilities and limitations in the planning and provision of local ALMPs in rural settings – perspectives from municipalities in northern Sweden
No full textMany welfare states in the Global North have decentralised active labour market programmes (ALMPs) from national to sub-national level. But how have these decentralisation processes, taking place against the backdrop of an urban bias, affected rural municipalities? The paper aims to improve the understanding of possibilities and limitations that arise in the planning and provision of ALMPs by local governments in rural settings, by exploring and analysing rural municipalities in northern Sweden. Through interviews with municipal ALMP planners, the paper illustrates challenges related to low economies of scale, the pace of national reforms, and support for groups with low employability
Dömd på indicier – en risk för rättssäkerheten? : En analys av indiciebevisningens betydelse och tillämpning i brottmål
No full textDevelopment and implementation of a Computer-Vision-System in a foundry
No full textDetta examensarbete har utförts i samarbete med företaget Baettr Guldsmedshyttan AB, ett företag som tillverkar segjärnskomponenter till vindkraftverk. Syftet med examensarbetet har varit att undersöka möjligheten att utveckla ett AI-baserat visionsystem anpassat till företagets behov inom sandgjutningsprocessen för att effektivt detektera ytdefekter. I dagsläget sker denna kontroll manuellt genom visuell inspektion, vilket medför variation i bedömningen och ergonomiska utmaningar. Ett automatiserat visionsystem kan bidra till en jämnare och mer tillförlitlig kvalitetskontroll samt en förbättrad och hållbar arbetsmiljö. Arbetet har fokuserat på olika detekteringsmetoder inom visionsystem, där anomalidetektering, en teknik som identifierar avvikelser från normaldata valdes som metod. Genom att använda mjukvaran Anomalib, utvecklad av Intel, och en träningsmetod baserad på AI-modellen PatchCore kunde ett system tränas på att urskilja defekta ytor från de normala. Resultatet visade att metoden är lovande, särskilt vid detektering av tydliga defekter som sprickor, medan mer subtila avvikelser som damm och sandrester kräver större datavariation och resurser. Metoden visade sig också fungera vid analys av rörligt material, som video. Studien visade att AI-baserade visionsystem har potential att förbättra kvalitetskontrollen i gjutprocesser, men fortsatt arbete krävs. Det innefattar mer omfattande datainsamling, utvärdering av andra ytdefekter, undersökning av alternativa AI-modeller samt investeringar i hårdvara, vilka är viktiga steg mot en mer hållbar framtid inom industrin.This thesis was conducted in a collaboration with Baettr Guldsmedshyttan AB, a company that manufactures ductile iron components for wind turbines. The purpose was to investigate the possibility of developing an AI-based vision system tailored to the company’s needs in the sandcasting process, with the goal of efficiently detecting surface defects. Currently, this inspection is done manually, leading to variation in assessment and ergonomic challenges. An automated vision system could provide more consistent and reliable quality control, while also contributing to a more sustainable work environment. The study focused on various detection approaches in computer vision, with anomaly detection being chosen as the primary method. This technique identifies deviations from normal data and was implemented using Intel’s Anomalib software and a training approach based on the AImodel PatchCore. The results indicate promising performance, especially for detecting distinct defects such as cracks, while more subtle anomalies like dust and sand residues require greater data variation and resources. The method also proved effective when analyzing moving material, such as video. The study concludes that AI-based vision systems hold significant potential for improving quality control in casting processes. However, further development is necessary, including more extensive data collection evaluation of additional defect types and AI models, and hardware investments, all of which are vital steps towards a more sustainable future in industrial manufacturing