263134 research outputs found
Sort by
Werkcondities voor leraren in de realisatie van krachtige leeromgevingen in de B-stroom: een meervoudige casestudy
No full textHet realiseren van ’krachtige leeromgevingen’ in de B-stroom is uitdagend voor leraren. Dit onderzoek identificeerde de werkcondities die het pedagogisch-didactisch handelen in de B-stroom kunnen ondersteunen. We onderzochten via een meervoudige casestudie van vier succesvolle scholen welke werkcondities leraren ondersteunen in het realiseren van krachtige leeromgevingen. De geïnterviewde leraren en de schoolleiding zien werkcondities zoals samenwerking, instructie-ondersteuning, innovatie en professionalisering als belangrijk en ondersteunend om in de B-stroom krachtige leeromgevingen te realiseren. Andere werkcondities, zoals professioneel vertrouwen en respect en voorbereidingstijd, faciliteren deze werkcondities. De resultaten bevestigen dat werkcondities systemisch zijn. De concretisering, het belang en de aanwezigheid van de werkcondities zijn schoolspecifiek. Deze condities kunnen andere scholen inspireren in het optimaliseren van de werkcontexten voor leraren, opdat ze krachtigere leeromgevingen kunnen verwezenlijken.status: Publishe
Naar flexibele en efficiënte hardware-architecturen voor machine learning - van modellering tot chipimplementatie
No full textDeep learning models power advances across computer vision, natural language processing, and other domains, but their significant computational and energy demands create deployment challenges. Cloud-based solutions consume substantial energy and raise privacy concerns, while edge deployment faces hardware constraints including limited resources and strict power budgets. This dissertation addresses these challenges through comprehensive hardware-software co-design optimization. The work spans hardware performance modeling to functional chip implementation, developing adaptable and energy-efficient architectures for deep learning acceleration. The dissertation presents hardware performance assessment frameworks, sparse neural processors, and specialized accelerators for emerging AI workloads including transformers and large language models. These optimization techniques enable sophisticated AI capabilities on resource-constrained platforms for seamless integration into everyday environments.status: Publishe
Active bystanders in the forwarding of sexting messages: Applying a theory of planned behavior in youth
No full textsponsorship: KU Leuven|C14/18/017status: Published onlin
Faagbiocontrole van gekke wortels en haar ecologische impact
No full textDue to the ever increasing prevalence of resistance to antibiotic, alternative treatments for infectious diseases are searched. Plant diseases are no exception and today, spraying antibiotics on the fiels is already forbidden in Europa. A promising alternative is phage biocontrol, in which virusses are used to treat bacterial plant pathogens. During this PhD, the potential of phages to treat hairy root disease will be investigated. This disease is caused by Agrobacterium rhizogenes biovar 1. the efficiency of the treatment will be assessed in greenhouses where tomatoes are cultivated. Apart from the efficience, also the impact of the treamtent in natural biome will be investigated, focussing n the virome. Also other factors that can affect the virome will be assessed.status: Publishe
Ontwarren van de genetische en moleculaire basis van de resistentie van rijst (Oryza sativa) tegen het Rice yellow mottle virus
No full textRice yellow mottle virus (RYMV) causes high losses to rice production in Africa. Thus, the region continues to grapple with fulfilling the demand for staple food, where it remains a net rice importer. Several sources of varietal resistance are available, but the emergence of virulent pathotypes that can overcome all known resistance alleles has been predicted.
The main aim of the research presented in this dissertation was to identify new breeding targets that modulate Rice yellow mottle virus resistance in high-yielding cultivated Oryza sativa (rice).
I present research on the genetic and molecular basis of rice resistance to RYMV using methods like genome-wide association studies (GWAS) and proteomics.
Two thousand rice accessions were screened for resistance against RYMV under two disease regimes—21 and 28 days after inoculation. The screening identified ten highly resistant accessions 28 days after inoculation. A genome-wide association analysis revealed four QTLs associated with RYMV resistance. Two QTLs, qRYMV11 and qRYMV12, were consistently identified 21 and 28 days after inoculation on chromosomes 11 and 12, respectively. A more in-depth analysis of these QTLs discovered that two genes had favourable alleles for RYMV resistance within the ten highly resistant accessions. Fine-mapping these QTLs will support future breeding programs by introducing the ten resistant accessions into elite cultivars.
Furthermore, an in-house proteomic approach called "Silica Acidic-based Phase Separation" was used to analyse changes in RNA-bound proteome during RYMV infection. RNA-binding proteins (RBPs) are critical as their perturbation influences the virus life cycle. The proteomic studies led to the discovery of 12 RBPs with increased abundance in response to RYMV infection. We envision that these RBPs represent candidate targets for functional validation of their roles in RYMV-infected plants.
Overall, the implications of these findings described here represent a significant step in understanding the genetic and molecular basis of RYMV infection in cultivated O. sativa. This sets a foundation for future molecular and genomics-assisted breeding applications in RYMV breeding programs.status: Publishe
Ventral mesh rectopexy: variations in technique and care process. A multicenter study
No full textAIM: The aim of this improvement collaborative is to explore the variation in care within and between Flemish hospitals in preoperative assessment, surgical indications, perioperative management and surgical technique for ventral mesh rectopexy (VMR). METHOD: This observational, cross-sectional multicentre study was performed in 14 Flemish hospitals. Twenty consecutive patients per hospital undergoing primary VMR in 2022 were included. Quality of care was assessed via predefined perioperative disease-specific quality indicators (QIs) by means of structured questionnaires. Data were collected from electronic patient files. RESULTS: A total of 280 patients were included. All patients were female and their mean age was 62 ± 14 years. Significant intra- and interhospital variation was observed in preoperative work-up, indications, operative technique and postoperative management. Total rectal prolapse was the indication for VMR in only 17.5% of the patients. The surgical approach was minimally invasive in all cases, with 40% via a robotic and 60% a laparoscopic approach. Fifteen per cent of patients had mechanical bowel preparation. All centres used a synthetic polypropylene mesh to perform a VMR, and in 85.6% (n = 238) of all patients a lightweight mesh was used. Diverging practices were noted as to type of mesh fixation to the rectum. In one third of patients a nonresorbable suture was combined with biological glue (n = 89, 31.8%). The overall mean length of stay was 2.1 (± 2.7) days. Only 3% of the procedures were performed as same day discharge, 47% of the patients remained for 1 day and 50% for ≥2 days. Only four patients were readmitted within 30 days after surgery. CONCLUSION: This study shows a significant variation in the perioperative management and surgical technique for VMR between hospitals, ongoing controversies and a lack of standardization. This collaborative can serve as a structured feedback tool to define minimum QIs and minimum outcome reporting parameters. Consensus building and adherence to evidence-based guidelines should reduce variation in care processes and lead to improved patient outcomes.status: Accepte
Robuuste Automatische Spraakherkenning en Gesproken Taalbegrip met Zwakke of Beperkte Supervisie
No full textThis dissertation investigates the development of robust speech models using limited or weak supervision. The primary focus is on addressing the challenges posed by variations in speech, such as accents, dialects, and background noise, which significantly impact the performance of Automatic Speech Recognition (ASR) and Spoken Language Understanding (SLU) systems. The scarcity of labeled speech data in many languages limits the ability to adequately represent all variations in speech through labeled examples. Therefore, this dissertation investigates techniques using unlabeled or weakly labeled data to develop robust speech recognition models and efficient SLU systems, proposing several novel methods and techniques to leverage approximately labeled and unlabeled speech data and to handle accented and noisy speech.status: Publishe
Oplossingen op basis van draadloze energieoverdracht voor afgelegen IoT-apparaten
No full textThis research explores innovative and novel approaches to provide energy to remotely deployed Internet of Things (IoT) devices. As the demand for continuous energy supply in remote environments increases, traditional battery-powered solutions face significant limitations regarding lifetime and sustainability. These limitations could be addressed by integrating and relying on wireless power transfer (WPT), for which a broad overview of technologies is included. Further, two unconventional solutions are studied: unmanned aerial vehicle (UAV)-based charging and radio frequency (RF) WPT systems. By investigating these solutions, the aim is to reduce battery capacity requirements or even achieve complete energy neutrality for extremely low-power IoT devices. The two proposed unconventional solutions are extensively elaborated upon, with the subsequent goal of extending device autonomy. This research employs analytical models and makes estimates based on simulations, finite element analysis and calculations to evaluate the performance of the proposed energy provisioning solutions. The feasibility of the discussed technologies are central to this work, with all associated challenges being examined to effectively implement these approaches. This research demonstrates that the UAV-based approach can exchange energy via a magnetic resonance coupling link. This loosely coupled technology offers advantages such as less stringent alignment requirements compared to closely coupled inductive power transfer. The proposed WPT implementation achieves efficiency levels of up to 40% and can support devices at ranges of up to several kilometres. This approach is particularly interesting for use cases that consume several joules of energy daily. Furthermore, it is shown that very low-power IoT devices, with daily energy requirements below 1 joule, can be powered using distributed near-field RF WPT implying distances to a few meters. The study of specific use case and experimental validations progress the technologies to potential valorisation. This research contributes to advancing the state of the art in energy provisioning for remote IoT applications.status: Publishe