Open Research Exeter - University of Exeter
Not a member yet
41213 research outputs found
Sort by
Yuanmingyuan’s Treasure: Biographies of the Qianlong Emperor’s Manuscript
No full text“Huangchao liqi tushi” (“皇朝禮器圖式”, Illustrations of Imperial Ritual Paraphernalia, “HCLQTS”
in abbreviation), a collection of illustrated manuscripts under commission of the Qianlong Emperor
(1711-1799), represents the emperor’s aspiration to create and codify a universal ritual system for the
imperium. In 1860, in a radical event of the Second Opium War between the Qing court and an
Anglo-French military alliance, Yuanmingyuan, once a splendid imperial palace complex, was burnt
down and spoiled. A collection of “HCLQTS” folios becomes one of numerous items of
Yuanmingyuan loot removed to Europe. Folios can be identified in Western museums, auctions, and
personal collectors’ hands. This historic object, involved in complex social relations across time and space, deserves in-depth
study. Regarding production, although completion is officially claimed in 1759, revisions continue
until 1772. Decades of work result in different colour-painting and printed editions. “HCLQTS”
content is complex, with six sections specifying ritual objects for particular ritual scenarios, including jiqi (“祭器”, sacrificial vessels), yiqi (“儀器”, scientific instruments), guanfu (“冠服”, costumes and accessories), yueqi (“樂器”, musical instruments), lubu (“鹵簿”, insignia), and wubei
(“武備”, arms and armours). Functions, meanings, and purposes are radically modified after the
1860 looting act, via changes to its human relations and social contexts. Kopytoff’s work on the
relationship and movement of objects between commodity status and singularity status, Appadurai’s
work on the social lives of things and object itineraries, and Bourdieu’s theory of distinction and
taste are used to frame the analysis, while recent discussions of contested heritage and decolonisation
of museums support the eventual conclusions. Considering the object as having social agency, a
dialectic relationship between human intentions and object autonomy is explored. I split “HCLQTS” social life into three stages: the indigenous, a turning point, and the European. In its indigenous life, “HCLQTS” emerges from Qianlong’s commission for his ritual reform, driven
by the emperor’s ritualisation of the imperium, his apperception of painting arts that is highly
resonant towards a European linear perspective, and an ideological construction of a Grand-Unity
China. The sacking of Yuanmingyuan in 1860 is the turning point that breaks the object unity and
mobilises it into discrete itinerants to Europe. Initially serving as a symbol of Qing sovereignty, the
European looting act starts diverting its career into a political symbol of European conquest over the
Qing empire, art commodities that, misrecognised, fail to achieve high economic value, and pieces of
mere ‘curiosities’ that are likely neglected. The fragmentation into discrete folios produces three trajectories. In the first, driven by different
auction houses’ and individual collectors’ interests, a batch of “HCLQTS” folios moves in and out of
the British art market. When held in collectors’ hands, it becomes singularised, meeting buyers’ personal experiences and emotions, differently from the auction house experience, where it is
commonly represented as an art good for profit-maximisation purposes. In the second trajectory, respectively, British Museum’s 1973 and the British Library’s 2002 exhibition, one single “HCLQTS”
folio serves as ideological instrument to craft British modernity, with colonial implications. Notably, the object history of British looting is silenced in both cases. In the third trajectory, as a display item
in the Victoria and Albert Museum’s permanent exhibition, the colonial echoes of another “HCLQTS”
folio’s social life take a more subtle form. While the looting history is still absent from public
representation, the object becomes a form of ‘chinoiserie’ that unfolds the institute’s prestige, while
indulging in racialisation of a Chinese Other, and evoking audiences’ fetishism towards Chinese
design and fashion. The thesis concludes with reflections for possible future research and museum
practice
Branching Out: How Academic Genealogy Networks Shape Academic Groups
No full textAcademic grouping and collaboration are fundamental to advancing scientific knowledge, as their composition directly influences productivity. However, accurately identifying the structure of these groups can be challenging because academics’ stated fields of study often do not align with their actual research output. While these discrepancies may seem minor, they can significantly impact an institution’s ability to make informed decisions, implement effective interventions, or provide appropriate incentives for the development of groups in emerging fields. Identifying these groups is essential for better assessing an institution’s performance and uncovering knowledge gaps. In this study, we employ an approach rooted in Network Science to analyse group composition, using an institution in Brazil as a case study. We further investigate the relationships between these groups and the academics’ research interests, examine the diversity of the groups in relation to their areas of expertise, and compare our findings with self-declared academic groups
Strategic influencers and the shaping of beliefs
No full textInfluencers, from propagandists to sellers, expend vast resources targeting agents who amplify their message through word-of-mouth communication. While agents differ in network position, they also differ in their bias: agents may naturally read articles with a particular slant or buy products from a certain seller. Absent competition, an influencer prefers targeting central agents and those biased against it. If agents are unbiased, competition leads to influencers targeting more central agents. However, when agents have heterogeneous biases and competition is intense, the incentive to deter one’s rival dominates. Influencers protect their base, targeting those with similar beliefs in equilibrium.</p
Developing physics-informed neural networks for virtual sensing in beams with moving loads
No full textPhysics-informed neural networks (PINNs) show promise for structural health monitoring and related applications since they have the potential to represent physical systems and processes by conforming to governing differential equations as well as physical constraints such as displacement and force boundary conditions. This paper addresses the challenge of developing PINNs that can predict the response of bridge structures under a moving concentrated load such as from the axle load for a vehicle. It also demonstrates the potential of PINNs for virtual sensing, i.e., predicting response parameters at locations where sensors may not be available. The proposed PINNs assume that the load moves at constant speed and uses the time at which the load enters the bridge to evaluate its position with respect to time. The study initially proposes a one-dimensional PINN that takes only response location as input and gradually increases dimensionality (number of inputs) to three by including the vehicle passage time and load magnitude as additional inputs. The study conducts a thorough sensitivity analysis of the PINN model’s parameters to understand their influence on model training and performance. The performance of the final PINN model is investigated for a real-world bridge girder for which monitoring data is available. The capacity of the PINN to predict strains, deflections and internal forces in locations that are not physically equipped with sensors is demonstrated, and strain predictions are observed to closely follow actual measurements from field monitoring.</p
Theoretical Understanding on Adversarial Robustness of Deep Neural Networks
No full textAs state-of-the-art (SOTA) Machine Learning (ML) techniques, particularly Deep Neural Networks (DNNs), become increasingly deployed across various industry sectors, concerns about their trustworthiness have been raised more frequently. The trustworthiness of DNNs can be undermined by adversarial examples—subtle input perturbations that are nearly imperceptible to humans but can deceive ML models. This can result in critical errors in applications such as cancer detection and traffic signal recognition. However, the underlying causes of adversarial examples still lack a comprehensive theoretical understanding.
To contribute to this theoretical understanding, we first investigate the layer-wise causes that influence the adversarial robustness of DNNs. We present a comprehensive theoretical framework that connects adversarial robustness to the Cauchy problem of its underlying Ordinary Differential Equation (ODE), allowing for the independent assessment of the robustness of neural network components, irrespective of attacking algorithms. Through two case studies, we illustrate the application of our theorem: first, we examine the robustness of commonly used bottleneck architectures and their variants, emphasizing the enhanced robustness of inverted residuals; second, we evaluate the defensive capabilities of Multihead Self-Attention (MSA) in Vision Transformers (ViTs), uncovering its inferior performance against strong adversarial attacks compared to convolutional neural networks.
In addition to focusing on individual layers, we investigate the effects of collaboration among them. We ask: Is there collaboration between layers against adversarial examples during gradient descent? To quantify this collaboration, we introduce a new concept called Collaboration Correlation (Co-Correlation), which is interpreted as the alignment of feature selections to maximise outputs for each layer. We examine the implicit bias of gradient descent for adversarial robustness and theoretically prove that gradient descent enhances the Co-Correlation between layers. Additionally, we observe differing behaviours in under- and over-parameterised neural networks: under-parameterised networks tend to promote Co-Correlation among layers to enhance performance, whereas the performance improvement of over-parameterised networks does not heavily rely on establishing such Co-Correlation.
With regard to generalisation and its connection to adversarial robustness, we establish the theoretical relationship between natural and adversarial risks. Building on this connection, we aim to answer the question: How does the cross-layer correlation of the weight matrices influence both natural and adversarial generalization errors? Under mild conditions, we provably show that natural risks monotonically increase with respect to the cross-layer correlations. However, the impact on adversarial robustness is more nuanced, depending on the mean value of the Jacobian of the weight matrices.
In this thesis, we investigate the internal structure of DNNs concerning adversarial robustness and generalization. We aim to understand how internal structures and their connections in neural networks influence adversarial robustness and its trade-offs with generalization. To this end, we propose a theoretical framework for analyzing these internal structures, offering a potential pathway for future research. From a practical standpoint, our exploration provides valuable guidance for the design and optimization of deep neural networks, such as identifying and replacing unstable modules to enhance adversarial robustness and employing tailored learning rate strategies for different layers to improve both performance and robustness. We believe that understanding the intrinsic properties of neural networks will also enhance the interpretability of large and complex models
Observer-based numerical schemes
No full textNumerical analysis and control theory are fundamental areas in engineering and
applied mathematics. This thesis explores three concepts from control theory used
to enhance classical numerical solvers. The proposed improvement integrates a
sampled-data Luenberger observer with a conventional numerical solver in a switched
system framework. The method employs the numerical solver when its updates are
sufficiently accurate and switches to using process samples to drive an observer when
they are not.
The switching mechanism is governed by an energy inequality based on a
Lyapunov function, potentially triggering sampling as needed. Stability proofs and
error estimates utilize input-to-state style stability techniques. This new numerical
scheme can handle step sizes significantly larger than those required for the stability
of the traditional numerical solver. Additionally, the hybrid approach of switching
between the sampled-data observer and the numerical solver can reduce the frequency
of sampling needed for accurate observer-based state estimation of the process.
In this context, this thesis has the following aims. Firstly, it combined any
numerical scheme with sampled-data Luenberger observer in a new hybrid scheme
based on switching conditions. The scheme uses the numerical scheme when scheme
updates are good enough but switches to an observer driven by process samples
when not. A Lyapunov function-based energy inequality determines switching. Thus,
the switching condition is central to the hybrid observer-based numerical scheme.
The idea underpinning this switching condition is to use a Lyapunov function for the
observer as an energy function for the Euler scheme. Loosely speaking, energy for the
observer’s solutions will decrease, and we only use the Euler scheme when its energy
also decreases. In this sense, the Lyapunov function for the observer becomes a Lyapunov function for the overall hybrid scheme. The switching condition partitions
the state space into sections or regions where we use Euler scheme and where we
use the observer. Depending on the system and the scheme’s parameters, the region
where Euler method is used may be large, small, or even null. Secondly, the aim is
to extend the generalized hybrid scheme with a higher-order approximation of the
Taylor exponential. We generalize the switched system by using Runge-Kutta. After
that, this hybrid ODE solver is constructed by combining the Euler and Luenberger
observer to switch from the numerical scheme to the observer when the numerical
scheme produces inadequate results. Underpinning our approach is a λ-tracking-based
sampled-data observer that invokes a λ dead zone. The resulting hybrid algorithm is
a time-stepping numerical scheme. The gains and sampling periods in the sampled-
data observer are tuned using a λ-tracking approach. Using a sampled-data observer
allows process measurements to be only available at some discrete times, while
adaptive tuning allows the gains and sampling times to adjust automatically to each
other rather than being subject to design. Finally, an alternative switching approach
is considered: switching from observer to Euler based on λ and μ strips
Meet Me on Tuesday (Screenplay): Place and Narratives of Ageing in Nigerian and South Korean Cinema
No full textThis thesis takes a comparative approach to exploring how older people are portrayed in contemporary Nigerian and South Korean cinema. In bringing together two distinct cinemas, the thesis argues that place is a fluid embodiment of culture which filmmakers negotiate in providing insights into how experiences of older characters from Nigerian and Korean cinema narratives are understood. The dominant literature on portrayals of older adults centres on Global North cinema depictions, with older people from African and Asian cinema underrepresented. This thesis identifies an evident gap and aims to contribute creatively and critically to ongoing research on representations of older people in African and Asian cinema.
The central question for this PhD is: How does the screenwriting practitioner and researcher negotiate experiences of ageing and place through the screenplay and in the comparative study of narratives from Nigerian and Korean world cinema?
The PhD is presented in two parts to explore this question: a 120-page feature length screenplay, E pàdé mi ni ojó Ìségun/Meet Me on Tuesday, and a critical thesis in four chapters, three of these chapters interrogate the notion of place and narratives of ageing in Nigerian and South Korean cinema. The fourth chapter explores the relationship between my creative practice and this critical research through a focus on how theories and practice developed the screenplay. A bridging chapter that describes how I have experimented with ideas and theories of ageing in the screenplay conjoins the two elements of this thesis. The screenplay, E pàdé mi ni ojó Ìségun/Meet Me on Tuesday, is both an output and a creative intervention to the questions asked in this thesis. In the critical component of the thesis, the first chapter examines intergenerational dynamics and place as a geographic entity through the lens of the older person in diasporic families in Lee Isaac Chung’s Minari (2020) and Andrew Dosunmu’s Mother of George (2013). In the second chapter, I explore the cultural productions of place with relation to the gendered nature of caring for older people in Lee Chang-dong’s Poetry (2010) and Ema Edosio’s Otiti (2022). Then, the past as a temporal place for older characters is interrogated through the life review theory in Yoon Je-kyoon’s Ode to My Father (2014) and Bolanle Austen-Peters’ biopic film Funmilayo Ransome Kuti (2024).
This research identifies the value of place in deepening our understanding of the older person across cinema cultures and more implicitly, my creative practice and screenplay constitute a form of care for and about older characters in cinema narratives, which I explore briefly in the critical thesis.</p
Preschool Teachers’ Perspectives and Experiences of supporting Children with Speech and Language Difficulties in Riyadh: Through the lens of Storytelling
No full textSaudi Arabia’s Vision 2030 (2016) emphasizes the right of every child in preschools to access quality educational opportunities, with a particular focus on early education, teacher training, and curriculum development. Despite these commitments, preschool teachers face persistent challenges, particularly in supporting children with difficulties such as speech and language difficulties (SLD). This study was conducted to explore these challenges by identifying their nature and examining potential ways to address them through two key objectives across two phases:
•Phase One: To evaluate the current situation in Riyadh preschools (public-independent, attachment, and private) regarding the identification of SLD, the provision of support, and the obstacles encountered. This was achieved through a mixed-methods approach, combining online questionnaires (158 respondents) and 13 semi-structured interviews.
•Phase Two: To investigate preschool teachers' approaches to supporting children with SLD during oral storytelling (OST) and digital storytelling (DST). This objective involved multiple qualitative case studies, including 21 classroom observations, 21 stimulated recall interviews, and 10 semi-structured interviews.
The findings from both phases were discussed using the Inclusive Pedagogy Approach in Action (IPAA) proposed by Florian and Spratt (2013), which emphasizes the development of educational practices that respond to individual differences without marginalizing or isolating learners, thereby promoting inclusive teaching.
Findings from the First Phase revealed that the participants expressed low confidence in identifying SLD. Participating teachers largely relied on their experiences and observations rather than diagnostic tools or standardized scales. Support mechanisms predominantly involved multimodal strategies combining images, sounds, and sensory materials, alongside social support through daily conversations, active listening, encouragement, and praise to foster a sense of belonging. However, 64.6% of participants reported no designated responsibility for children with SLD, citing inadequate knowledge (63.3%) and lack of relevant training (62%). Disparities in salaries between educators with degrees in early childhood education and special education further hindered some teachers’ full engagement in supporting children with SLD.
The Second Phase explored teachers’ approaches during OST and DST. Findings showed inconsistency in OST practices: while some participants implemented inclusive strategies such as simplifying stories and one-on-one teaching, others were constrained by systemic challenges like large class sizes and limited resources. DST, on the other hand, emerged as a promising approach, offering flexibility through small group activities that enhanced engagement and comprehension for children with SLD. However, resource limitations and the additional workload required for DST preparation hindered its broader implementation.
Participants consistently highlighted a lack of training, resources, and institutional support as significant barriers to effective support during both OST and DST. Addressing these challenges requires measures such as reducing class sizes, providing comprehensive guides tailored to OST and DST, and offering targeted professional development focused on supporting children with SLD. The findings from this study offer valuable insights and practical recommendations for optimizing inclusive education in Saudi preschools.</p
The experiences of bedside nurses delivering an intensive care sedation study: a process evaluation within the A2B trial
No full textBackground:This process evaluation explored delivery of a complex sedation intervention within the Alpha-2 Agonists for Sedation to Produce Better Outcomes from Critical Illness (A2B) trial, which compared dexmedetomidine- and clonidine-based sedation with propofol (usual care). All groups targeted lighter sedation levels. The objective was to understand bedside nurses’ experiences delivering the interventions and identify factors influencing protocol adherence and implementation.Methods:A qualitative study using semi-structured interviews was conducted with intensive care unit (ICU) staff (consultants, bedside and research nurses) from A2B trial sites. Thematic analysis explored how participants experienced and delivered trial interventions, with particular focus on bedside nurses’ abilities to manage sedation in line with the protocol.Results:Nurses with greater ICU experience described more confidence and adaptability in using alpha-2 agonists, while less experienced staff required support due to limited familiarity with lighter sedation. Hesitancy to up-titrate alpha-2 agonists was common, driven by concerns about bradycardia and hypotension. Reluctance to down-titrate propofol was shaped by fears of agitation and self-extubation. Deep sedation norms, especially amongst nurses trained during the COVID-19 pandemic, further hindered protocol adherence. Research nurses were instrumental in supporting implementation and bridging knowledge gaps.Conclusion:Despite all three trial groups targeting lighter sedation, nurse confidence, safety concerns, and ingrained cultural practices limited adherence to alpha-2 agonist-based protocols. Addressing these barriers through training, support, and cultural change will be essential for future trials and practice shifts involving lighter sedation strategies in ICU.Trial registration number:ClinicalTrials.gov NCT03653832 https://clinicaltrials.gov/study/NCT03653832</p
The future of (environmental) history: A roundtable discussion
No full textIn April 2023, eighteen scholars from nine different subjects representing the humanities, natural and social sciences came together for a one-day workshop at St John's College, Durham. Despite our differences, all had one aim: the study of past environmental change and its effects on human societies. Talking across disciplinary divides, we discussed what environmental history is, how it may or may not contribute to tackling the climate crisis, and the problems of sources, scale and temporality. This article collects select conversations into a roundtable format split into four areas: scale, time and space, interdisciplinarity, and the future of environmental history. We argue that environmental history is more usefully understood not as a distinct sub-field of history, but as an interdisciplinary meeting place for innovative collaboration. This also presents a model for future research aimed at tackling the climate crisis at higher education institutions.</p