157082 research outputs found
Sort by
AI-Augmented Creativity in Learning Analytics
No full textIn an era where AI technology is rapidly redefining the boundaries of knowledge, creativity, and learning, learning analytics is no longer limited to collecting behavioral data. It has become a new platform for discovering hidden patterns, understanding the learner experience more deeply, and designing personalized learning paths. AI-augmented creativity, as an emerging concept, has found a key position in the transformation of education and learning design. This approach not only helps to analyze educational data more accurately but also enables the creation of innovative and inspiring solutions that were previously beyond the power of humans or traditional algorithms.AI-Augmented Creativity in Learning Analytics explores the innovative pathways for integrating learning analytics with AI. This book offers a fresh perspective on designing personalized, engaging, and data-informed learning journeys for the future of education. Covering topics such as education, AI, and learning analytics, this book is an excellent resource for researchers, educators, instructional designers, content developers, and educational policymakers seeking to better understand the intersection of data, creativity, and technology
AIS underrepresents vessel traffic in Scotland's Marine Protected Areas
No full textMaritime traffic poses a variety of risks to both the marine environment and marine wildlife. To quantify and predict risk, accurate data on the distribution and densities of vessel traffic is required, yet currently there is no single data type that captures all vessel traffic. Most commonly, AIS (Automatic Identification System) vessel tracking data is used, despite awareness that AIS data does not fully capture all vessels present. Therefore, evaluations using only AIS likely underestimate the potential impacts. To estimate the scale of underestimation, vessel presence within six of Scotland's Marine Protected Areas (MPAs) were recorded during >1800 h of land-based and at-sea surveys, and compared with AIS data collected from a network of receivers deployed around Scotland. Non-AIS vessels were present within MPAs during 62 % of the surveyed period, with 64 % of vessels sighted not broadcasting AIS. AIS transmission rates varied between MPA, season and vessel type. Given that AIS data is the most commonly used data type for quantifying vessel activity and predicting associated impacts, consideration must be given to the volume of vessel traffic not represented within AIS datasets, particularly within MPAs. Underestimation of actual vessel traffic is likely leading to insufficient management or mitigation efforts within areas designated for protection
NLP verification:towards a general methodology for certifying robustness
No full textMachine learning has exhibited substantial success in the field of natural language processing (NLP). For example, large language models have empirically proven to be capable of producing text of high complexity and cohesion. However, at the same time, they are prone to inaccuracies and hallucinations. As these systems are increasingly integrated into real-world applications, ensuring their safety and reliability becomes a primary concern. There are safety critical contexts where such models must be robust to variability or attack and give guarantees over their output. Computer vision had pioneered the use of formal verification of neural networks for such scenarios and developed common verification standards and pipelines, leveraging precise formal reasoning about geometric properties of data manifolds. In contrast, NLP verification methods have only recently appeared in the literature. While presenting sophisticated algorithms in their own right, these papers have not yet crystallised into a common methodology. They are often light on the pragmatical issues of NLP verification, and the area remains fragmented. In this paper, we attempt to distil and evaluate general components of an NLP verification pipeline that emerges from the progress in the field to date. Our contributions are twofold. First, we propose a general methodology to analyse the effect of the embedding gap - a problem that refers to the discrepancy between verification of geometric subspaces, and the semantic meaning of sentences which the geometric subspaces are supposed to represent. We propose a number of practical NLP methods that can help to quantify the effects of the embedding gap. Second, we give a general method for training and verification of neural networks that leverages a more precise geometric estimation of semantic similarity of sentences in the embedding space and helps to overcome the effects of the embedding gap in practice.</p
Robust feedback control of collisional plasma dynamics in presence of uncertainties
No full textMagnetic fusion aims to confine high-temperature plasma within a device, enabling the fusion of deuterium and tritium nuclei to release energy. Due to the very large temperatures involved, it is essential to isolate the plasma from the device walls to prevent structural damage and the external magnetic fields play a fundamental role in achieving this confinement. In realistic settings, the physical mechanisms governing plasma behavior are highly complex, involving numerous uncertain parameters and intricate particle interactions, such as collisions, that significantly affect both confinement efficiency and overall stability. In this work, we address particularly these challenges by proposing a robust feedback control strategy designed to steer the plasma towards a desired spatial region, despite the presence of uncertainties. From a modeling perspective, we consider a collisional plasma described by a Vlasov-Poisson-BGK system, which accounts for a self-consistent electric field and a strong external magnetic field, while incorporating uncertainty in the model. A key feature of the proposed control strategy is its independence from the random parameter, making it particularly suitable for practical applications. A series of numerical simulations confirms the effectiveness of our approach and demonstrates the ability of external magnetic fields to successfully confine plasma away from the device boundaries, even in the presence of uncertain conditions
Changing EAP assessment practices in the age of generative artificial intelligence: The case of Scottish higher education institutions
No full textThe impact of generative artificial intelligence (GenAI) on higher education has been widely discussed since the public release of ChatGPT-3.5 in late 2022. However, there has been little empirical research on changes in English-for-Academic-Purposes (EAP) assessment practices in response to GenAI. This qualitative case study intends to fill this gap by examining how Scottish universities changed EAP assessments in response to GenAI, how effective those changes were perceived by EAP academics, and what recommendations EAP academics offered for future assessment practices. Data were collected from six semi-structured interviews conducted with EAP academics at five Scottish universities in mid-2024 and thematically analysed. The findings reveal that while substantial changes in assessment task design were limited, modifications to task requirements (e.g., GenAI declarations, context-specific prompts) and grading practices were more common. Moreover, our participants expressed scepticism about the effectiveness of some changes (e.g., AI use declarations) but positively perceived others (e.g., the use of context-specific questions, spontaneous speaking tasks, and named marking). As for their recommendations, the participating EAP academics generally advocated authentic and innovative tasks, such as portfolio-based assessment, reflections, multimodal projects, and GenAI output evaluation over reverting to traditional exams while simultaneously highlighting issues with workload and learning outcomes. The study implies a need for clearer institutional guidance, ongoing professional dialogue, and support for experimentation with GenAI-integrated assessment design in EAP contexts
Ordering groups and the Identity Problem
No full textThe Identity Problem — deciding if the subsemigroup generated by a given finite set of elements of a group contains the identity element — is shown in this paper to correspond, for certain classes, to decision problems about ordering groups. Notably, the Identity Problem for a torsion-free nilpotent group corresponds both to the problem of deciding if a given finite set of elements extends to the positive cone of a left-order on the group, and to the Word Problem for a related lattice-ordered group.A new (independent) proof is given of the decidability of the Identity and Subgroup Problems for every finitely presented nilpotent group (initially proved by Shafrir in 2024), establishing also the decidability of the Word Problem for a family of lattice-ordered groups. In contrast, it is shown that the related Fixed-Target Submonoid Membership Problem is undecidable in nilpotent groups.Decidability of the Normal Identity Problem (with ‘subsemigroup’ replaced by ‘normal subsemigroup’) for free nilpotent groups is established using the (known) decidability of the Word Problem for certain lattice-ordered groups. Connections between orderability and the Identity Problem for a class of torsion-free metabelian groups are also explored
Enhancing energy transport utilising permanent molecular dipoles
No full textWe study exciton energy transfer along a molecular chain while accounting for the effects of permanent dipoles induced by charge displacements in the molecular orbitals. These effects are typically neglected as they do not arise in conventional atomic quantum optics; however, they can play an important role in molecular systems. We further demonstrate how permanent dipoles can preferentially arrange energy eigenstates to support excitation transport, such as by creating energetic barriers and adding structure to the eigenspectrum. Putting all this together, we show how permanent dipoles can enhance the ability of the molecular chain to support excitation transport across a range of environmental and system configurations
NanoCoating Preparation to Improve Heat Dissipation of a Heat Sink Inside an Enclosure for Power Electronic Devices
No full textHeat sinks dissipate heat from electronic components, and the increase in heat generation owing to technological advancements has prompted researchers to improve heat sink efficiency. The present study aims to improve heat sinks by high-emissivity nanocoating where the coating is prepared using nanoparticles CuO and MWCNT at a rate of 6% in half a liter of Acrylic resin and solvents Xylene and Butyl acetate at a rate of 30%. After coating the heat sink, the emissivity was examined and it was (0.963) while it was before coating (0.202). The heat sink is examined inside a cubic cavity with a right surface containing heaters that give temperature at the same value as the thyristors (58.5°C, 90°C, and 112.5°C) and a cold left surface (30°C). The temperatures at the tip of each fin are measured before and after coating when they change with time and at a steady state. The results showed that the nanocoating significantly reduced the temperature compared to the uncoated condition with the improvement percentage at a heater temperature of 58.5°C ranging from 10% to 15% at 90°C ranging from 24% to 34% and at 112.5°C ranging from 23% to 35%. It is concluded that the nanocoating showed great effectiveness in improving the performance of the heat sink at all temperatures, but the maximum effectiveness was at high thermal loads
A Dynamics-Based Method for Determining the Local Finite Mobility of Single-loop Spatial Mechanisms
No full textThis paper proposes a method for calculating the local finite mobility of single-loop spatial mechanisms based on modal analysis. Using spanning tree-based multibody dynamics, the single-loop spatial mechanism is modeled as a tree-like kinematic chain with serial chains closed by the constraints represented by a spring force model. The dynamic model is linearized using Taylor expansion. The stiffness matrix is then yield. The correspondence between vibrating/non-vibrating generalized coordinates and the nonzero/zero eigenvalues in the stiffness matrix is clarified. The mobility of the single loop spatial mechanism is then determined by the number of zero eigenvalues in the stiffness matrix. The method is then validated and analyzed by calculating the DOF of Sarrus mechanism, Bennett mechanism, 3-mode 7R mechanism, a mechanism with special parameters and a variable-DOF 8R mechanism. One contribution is that this work enhances spanning-tree-based dynamic modeling by analyzing joint selection strategies and introducing spring forces to replace kinematic constraints. The Other contribution is that based on the linearized model, a modal analysis framework is established to determine the mobility of single-loop spatial mechanisms.</p
A Lightweight Model LGCSPNet for Sitting Posture Risk Management Applications
No full textCurrent methods for sitting posture recognition typically follow a pipeline involving keypoint extraction and skeleton graph construction, followed by pose classification using Convolutional Neural Networks (CNNs) or Vision Transformers (ViTs). However, CNNs struggle to model long-range dependencies among keypoints, whereas ViTs suffer from high computational costs. Moreover, both approaches tend to introduce redundancy during feature modeling. To improve efficiency, some studies have explored direct classification using keypoint coordinates, but these methods often fail to balance high accuracy with computational efficiency. To this end, this paper proposes a new model LGCSPNet with lightweight graph convolution modules (LGC) and a contrastive learning module. Firstly, LGC enables efficient full-keypoint communication by shifting features across keypoint channels, allowing each keypoint to access global context at minimal computational cost. Building on this, LGC enhances sitting posture detection by computing 3D attention weights via a parameter-free energy function with a closed-form solution, enhancing feature learning for posturally significant keypoints. The contrastive learning module enhances differentiation between similar postures in different categories by strategically selecting feature samples. Experiments on public human posture datasets and our custom sitting posture dataset show that LGCSPNet has only 0.097M parameters while achieving a 99% recognition rate. It surpasses existing models in terms of parameter quantity and accuracy. Guided by ergonomic metrics, our model enables posture correction and mitigates long-term sitting-related injuries
Access Repository Dashboard
Do you manage Heriot Watt Pure? Access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard!