Open Research Exeter - University of Exeter
Not a member yet
41213 research outputs found
Sort by
Microwave quantum heterodyne sensing using a continuous concatenated dynamical decoupling protocol
No full textBy sequentially recording the phase of an AC signal relative to an external clock, quantum heterodyne schemes have recorded MHz and GHz signals with Fourier-limited precision. However, in systems with large inhomogeneous broadening, existing heterodyne protocols provide limited protection of the spin coherence, impacting amplitude sensitivity. Here, we use a continuous microwave scheme that extends spin coherence towards the effective T2≈12T1 limit and resolves the frequency, amplitude and phase of MHz to GHz magnetic fields. In an ensemble of boron vacancies in hexagonal boron nitride the scheme achieves an amplitude sensitivity of η≈3−5μT/Hz and phase sensitivity of ηϕ≈0.076rads/Hz. We demonstrate that the scheme is compatible with quantum heterodyne detection, recording a GHz signal with a resolution </p
Mutual Knowledge Distillation-based personalized Federated Learning for smart edge computing
No full textFederated Learning (FL) is a privacy-preserving machine learning paradigm that aims to train a global model using heterogeneous data across clients, which are typically consumer electronic devices such as smartphones, smart vehicles, and smart home appliances. As the global model may not be optimal for individual clients with unique behaviours, Personalized Federated Learning (PFL) was proposed to enable clients to adapt the global model to their specific needs and preferences. Nonetheless, due to the variance in data distributions across clients, the global model utilized in PFL may ‘catastrophically forget’ the knowledge gained in previous communication rounds, thereby leading to unstable performance. To address this challenge, we propose FedMKD, a novel PFL algorithm based on Mutual Knowledge Distillation (MKD) and elastic weight consolidation (EWC). FedMKD enhances the global model’s performance by addressing ‘catastrophic forgetting’ through EWC regularization, while enabling clients’ local models to effectively leverage the global model’s knowledge via MKD. Moreover, we apply uniform/exponential quantization methods to compress model parameters to decrease communication overheads. Experimental results demonstrate that FedMKD outperforms several key PFL baselines, FedMKD can also significantly reduce communication overhead while preserving its performance using suitable compression techniques, making it highly suitable for resource-constrained smart edge computing environment.</p
NiS nanoparticle-MoS<sub>2</sub> nanosheet core-shell spheres: PVP-assisted synthesis and efficient electrocatalyst for hydrogen evolution reaction
No full textThe synthesis of an effective, robust, and cost-efficient electrocatalyst for H2 through water electrolysis is crucial for developing clean and sustainable energy technology. Herein, we report polyvinylpyrrolidone-assisted synthesis of NiS nanoparticle-MoS2 nanosheets core-shell (NiS@MoS2) spheres as an efficient electrocatalyst for the hydrogen evolution reaction (HER) using the hydrothermal method. The polyvinylpyrrolidone controls the size and morphology of the NiS@MoS2 sphere, and it converts into N-doped carbon (NC) as a conductive material during the calcination process to form NiS@MoS2/NC spheres. The change in MoS2 concentration (NiS@MoS2/NCx) of different catalysts significantly influences the material, size, and electrocatalytic performance. NiS@MoS2/NC50 exhibits excellent HER performance, showing a small Tafel slope of 56.7 mV dec-1 and good stability with an overpotential of 189 mV at a current density of 10 mA cm-2. HER performance is believed to be enhanced by the synergistic interaction between NiS nanoparticles and MoS2 nanosheets at the heterointerface. Furthermore, NC enhances the conductivity of the NiS@MoS2/NCx composites to enhance HER performance. Density Functional Theory simulations indicate that the best catalytic site of NiS@MoS2 is Mo, where optimal electron transfer occurs between MoS2 nanosheets and NiS nanoparticles. This electron transfer promotes H+ adsorption, reducing Gibbs free energy (ΔGH* ≈ 0.04 eV) and requiring a work function of 4.81 eV, thereby boosting HER performance. This work will open a low-cost approach for synthesis of non-precious core-shell electrocatalysts with high activity and stability for HER.</p
An archaeology of risk: a bibliometric analysis of risk research
No full textThis study presents a bibliometric analysis of two journals central to the field of risk research—Risk Analysis and Journal of Risk Research to outline what has been accomplished, evaluate how the field has developed, and uncover potential future avenues of research focus. Conducting a bibliometric analysis of journal article keywords collected from the Scopus database between 1981 and 2023, we evaluate the risk research field using quantitative methods, VOSViewer visualisation software and a reflexive inductive and deductive thematic analysis. Using these methods, we analyse and map the entire risk research field between 1981 and 2023, representing the most complete bibliometric analysis of keywords to date. We also examine trends in keyword use over time and its co-occurrence or clustering with other keywords, focusing particularly on changes in the use of keywords that signify (1) core risk research domains; (2) risk theories and concepts, and (3) hazard topics. We find, among other things, that while there is growth in the number of publications and their diversity, these changes are uneven between different keywords from the three keyword conceptualisations. They are often driven by socio-political context, reactiveness to hazardous events, research process, funding, and/or publisher-related drivers. We conclude with implications for various groups engaging with risk research and recommendations for future research examining the history of the discipline.</p
Knowing oneself with and through AI: from self-tracking to chatbots
No full textThis chapter examines how algorithms and artificial intelligence are
transforming our practices of self-knowledge, self-understanding, and self-narration.
Drawing on frameworks from distributed cognition, I analyse three key domains
where AI shapes how and what we come to know about ourselves: self-tracking
applications, technologically-distributed autobiographical memories, and narrative
co-construction with Large Language Models (LLMs). While self-tracking devices
promise enhanced self-knowledge through quantified data, they also impose
particular frameworks that can crowd out other forms of self-understanding and
promote self-optimization. Digital technologies increasingly serve as repositories for
our autobiographical memories and self-narratives, offering benefits such as detailed
record-keeping and scaffolding during difficult periods, but also creating
vulnerabilities to algorithmic manipulation. Finally, conversational AI introduces new
possibilities for interactive narrative construction that mimics interpersonal dialogue.
While LLMs can provide valuable support for self-exploration, they also present risks
of narrative deference and the construction of self-narratives that are detached from
reality.</p
Trustworthy Machine Learning for High Assurance Systems
No full textDeploying neural networks in safety-critical applications requires mathematical guarantees about their behaviour that cannot be provided through empirical testing alone. While neural networks have demonstrated remarkable performance across diverse domains, their susceptibility to adversarial examples and lack of formal specifications present challenges for high-assurance systems where predictable behaviour is essential. This thesis presents a comprehensive formal neural network verification framework using interactive theorem proving in Isabelle/HOL.
First, we establish a complete formalisation of interval arithmetic for neural networks. Second, we develop three complementary formal representations of neural networks (directed graph, list-based, and matrix-based sequential models), each optimised for different verification scenarios with proven equivalences that enable conversion between the representations. Third, we prove the inclusion isotonicity, interval extension, and Lipschitz continuity of neural network computations.
Our verification methodology provides three approaches. Closed-form verification produces exact mathematical definitions through symbolic execution. Interval analysis provides a scalable approximation with soundness guarantees. Interval refinement combines both approaches through subdivision and refinement, which offers precision that converges to arbitrary accuracy.
Our tooling infrastructure enables a complete workflow of formal machine learning verification. An automated import mechanism translates trained TensorFlow networks into formal representations while automatically establishing various mathematical properties required for verification.
Our approach is demonstrated through different case studies. Verification of a digit recognition network provides a comparative analysis of all three verification approaches, while the ACAS Xu airborne collision avoidance system case study demonstrates scalability to safety-critical applications.
Key contributions include: (1) the first comprehensive neural network verification framework in an interactive theorem prover capable of handling networks trained with standard machine learning frameworks; (2) the development of novel verification methods that allow for formal analysis and reasoning about complex network properties; (3) demonstrated practical applicability through successful verification of safety-critical benchmarks.</p
Left-behind or Special Educational Needs, How They Interact to Affect Inclusion in Chinese Rural Areas
No full textThis study investigates the interplay between Left-Behind (LB) status and Special Educational Needs and Disabilities (SEND) in affecting inclusion in rural Chinese areas. The study is rooted in the theoretical context of inclusive education and the specific challenges faced by LB children and those with SEND. The research objectives include determining the prevalence and distribution of SEND among rural children, assessing the impact of LB and SEND statuses on inclusion, and exploring the interaction between these statuses.
The methodology adopted a mixed-methods approach, grounded in a pragmatic philosophical stance and utilising the bioecological framework as a valuable lens for understanding inclusion. The research design consisted of three phases: an initial teacher questionnaire to identify children with LB and SEND; a student questionnaire assessing dimensions of social inclusion, including peer acceptance, rejection, belonging, and bullying; and in-depth case studies involving interviews with teachers and caregivers. Quantitative data were analysed using descriptive statistics, chi-square tests, independent samples t-tests, two-way ANOVAs, and Pearson correlation coefficients. Qualitative data were examined thematically to contextualise and expand upon the statistical findings.
Findings indicate that 22.8% of children were identified with SEND, with no statistically significant differences between LB and non-LB groups in either prevalence or type. LB children were found to be more accepted by peers, contrary to prevailing assumptions of social marginalisation, while children with SEND experienced marked social challenges, including lower acceptance, diminished sense of belonging, and increased rates of bullying.
The research further reveals that children with both LB and SEND status demonstrated better social acceptance than their non-LB counterparts with SEND, suggesting that shared experiences within rural communities may mitigate certain social barriers. Nevertheless, the compounded difficulties for LB-SEND children were evident, particularly in relation to academic support and family involvement. Structural limitations in rural education, such as inadequate teacher training, limited specialist provision, and a dominant emphasis on academic performance, were identified as significant impediments to meaningful inclusion.
By drawing on theories of social inclusion, child development, and resilience and situating the analysis within the socio-cultural context of rural China, the study contributes to existing literature by challenging deficit-based perspectives. It highlights the need for contextually responsive educational practices and policies that acknowledge the complex, intersecting identities of rural children. The findings highlight the need for integrated, locally tailored strategies to support inclusion, particularly for those facing multiple forms of disadvantage.</p
PySimPace v2.0: An Easy-to-Use Simulation Tool with Machine Learning Pipelines for Realistic MRI Motion Artifact Generation
No full textMotion artifacts in structural and functional magnetic resonance imaging (MRI) pose a significant challenge for both clinical use and machine learning (ML)-based image analysis. Existing ML approaches for artifact correction require paired clean and corrupted datasets, which are difficult to acquire. We present py-simpace, an open-source, pip-installable MRI motion artifact simulation toolkit with native ML integration. py-simpace supports structural MRI and functional MRI (fMRI) simulation, offering configurable k-space and image-space motion, ghosting, Gibbs ringing, and physiological noise. It provides an end-to-end pipeline with a ready-to-use PyTorch Dataset interface for ML training. We describe the design of py-simpace v2.0, compare it with existing tools, and demonstrate its utility for robust artifact correction model development.</p
Strategic Remote-Operated Valve placement in looped water transmission networks: Risk-based and hydraulic impact–driven approach
No full textTimely isolation of burst locations in water transmission systems is critical for mitigating extensive water loss (WL) and minimizing disruptions in water supply. However, delays in postfailure action time such as decision-making and manual valve operations can lead to substantial infrastructure damage and prolonged service interruption. This study proposes a novel, field-informed framework that integrates machine learning, graph theory, and pressure-driven hydraulic simulation to quantify the benefits of retrofitting existing valves with remote operation and prioritizes installation location. A machine learning-based probability of failure (PoF) model, trained on long-term asset data, is used to simulate realistic burst scenarios. For each failure case, optimal isolation and interconnection valve sets are identified through graph-based analysis combined with hydraulic validation. Field-based manual operation delays are estimated using the Open-Source Routing Machine (OSRM), incorporating actual road network data to reflect true response times. The performance of remote-operated valves (ROVs) is assessed using two quantitative indicators: WL (as a proxy for cascading indirect impacts) and water supply shortage (WSS; direct impact). By comparing manual and ROV-enabled responses, the framework identifies the optimal ROV deployment for each segment, depending on specific operational conditions such as redundancy and supply availability. The analysis shows how the effectiveness of ROVs varies depending on network resilience, as well as how ranking valve sets by marginal benefits can support efficient investment decisions. Overall, this approach enhances burst response effectiveness by integrating proactive planning with postfailure interventions and supports practical decision-making in complex transmission systems.</p
How Can Traditional Culture And Creativity Be Symbiotic In Chinese Higher Education In Dance?
No full textInternationally, researching creativity in dance has high potential research value. Especially in the field of higher dance education, how to cultivate students' creativity has been one of the focuses of dance studies. However, both internationally and in China, there has been little research on creativity in folk dance in Chinese higher dance education. This thesis therefore investigates the practical teaching of folk dance in Chinese higher dance education, particularly focusing on how traditional culture and creativity coexist in Chinese higher dance education. Rather than relying exclusively on Western theories and philosophies to study Chinese folk dances, this study combines socio-constructivism, interpretivism, and phenomenology of perception, and the inclusivism of Chinese philosopher Zhuangzi, to provide the ontological, epistemological and axiological framework for the research. Through such a philosophical framework, this study can provide a multidimensional conceptualisation of dance creativity from the perspective of Chinese folk dance in higher education (HE). This study therefore builds a bridge between the study of Chinese and Western dance creativity.
Using a qualitative research approach, this study conducted observations, semi-structured interviews and focus groups with folk dance teachers, students and dance experts in three different Chinese dance universities, and thematically analysed the data using Nvivo software, in order to explore the understanding of creativity in Chinese higher dance education from the participants’ perspectives.
The research found that creativity was conceptualised in varied ways in the field of Chinese folk dance, where creativity was categorised into four concepts: novelty, innovation, originality and high level of originality. Furthermore, after combining the analysis of visual and textual data, the research found that teachers' and students' individual creativity was presented in different ways. By combining the findings of four sub-research questions, this research concludes that creativity is defined in the area of folk dance specialisation in Chinese higher dance education as the unique understanding and manifestation of dance by teachers, students and dance experts, which is derived from life and the heritage of traditional Chinese culture. Students and teachers are encouraged to develop their own understanding of dance while practising Chinese folk dance, in order to develop creativity through their own bodies and unique teaching approaches. Using traditional culture as the foundation to develop this creativity in folk dance is the key to the co-existence relationship between the two.
Overall, this study emphasises the diversity of creativity definitions in Chinese dance HE stemming from different socio-cultural and educational backgrounds. In so doing, this study starts to fill the gap regarding the current lack of a diverse theory of dance creativity in Chinese HE. It addresses concerns other than individual creativity, with the diversity and complexity of creativity concepts having implications for the development of understandings of creativity in Chinese HE dance.
The data also demonstrates that, in practice, people's understandings of creativity can be inconsistent and complicated. This research therefore demonstrates the complexity of the understandings of creativity in Chinese dance HE. It suggests that it is a conceptual area in need of community exploration which might grow through future nuanced exchanges between Chinese and Western academics and their developing inter-related conceptions of creativity.</p