53417 research outputs found
Sort by
From Symbol to Asset: The Trade Mark–Investment Interface and the Future of Regulatory Sovereignty
Get PDFGeneral dental practitioners’ training and experience in clear aligner treatment in England: a mixed-methods study
No full text
The Nottingham consensus on dementia risk reduction policy: recommendations from a modified Delphi process.
No full textTranslation of evidence about dementia risk and its reduction into effective, equitable public health policy is a major challenge. To address this challenge, the National Institute for Health and Care Research Policy Research Unit in Dementia and Neurodegeneration at Queen Mary University of London (DeNPRU-QM) convened a multidisciplinary panel of 40 experts from across England, with diverse lived, academic, clinical, policy and advocacy experience, at various career stages, and of diverse gender and ethnicity, to develop actionable policy recommendations for dementia risk reduction. Through a 2-day in-person workshop and a subsequent three-round modified Delphi survey, the panel evaluated and refined statements on dementia prevention. The panel achieved consensus on 56 recommendations in four domains: public health messaging, individual-level interventions, population-level interventions and research commissioning. A key priority across all domains was the need to consider and address health inequalities so that prevention efforts do not exacerbate existing disparities. Our recommendations provide policymakers with a robust foundation for designing and implementing an evidence-based dementia prevention strategy in England and provide guidance that can inform approaches in other countries and contexts. By prioritizing clear communication, targeted intervention and sustained research investment, the recommendations can help to address structural inequities and advance dementia risk reduction. Ongoing cross-sector advocacy will be crucial in driving policy adoption and implementation
Simone Weil and Judaism
No full textThis essay argues that Simone Weil’s fraught relationship with Judaism reveals the ethical and political dimensions of her wider philosophy. It examines Weil’s denunciation of Jewish theology—particularly her critiques of idolatry, nationalism, and the worship of power—and situates these within Weil's broader rejection of force and collective identity. The discussion traces how Weil’s views have provoked accusations of antisemitism while also inspiring nuanced reassessments by thinkers such as Levinas, Chenavier, and Rose. The essay contends that, although Weil’s judgments are often partial and severe, they expose enduring tensions between religion, nationalism, and universal justice. In its final sections, the piece reconsiders Weil’s relevance in the context of Israel’s war on Gaza, arguing that her thought unsettles assumptions about the relationship between Judaism and Zionism. Ultimately, the essay presents Weil’s radical universalism as a challenge to all forms of moral exclusion and political idolatry
Leverage cross-domain variations for generalizable person ReID representation learning
No full textGeneralizable person ReID has significant practical value in challenging the fragile i.i.d. assumption by learning a domain-generalizable person representation applicable to out-of-distribution test samples. Existing methods explore feature disentanglement to learn a compact generic feature space by eliminating domain-specific knowledge. Such methods not only sacrifice discrimination in target domains but also limit the model's robustness against per-identity appearance variations across views, which is an inherent characteristic of ReID. In this work, we formulate Cross-Domain Variations Mining (CDVM) to simultaneously explore explicit domain-specific knowledge while advancing generalizable representation learning. Our key insight is that cross-domain style variations need to be explicitly modelled to represent per-identity cross-view appearance changes. This approach retains the model's robustness against cross-view style variations that can reflect the specific characteristics of different domains whilst maximizing the learning of a globally generalizable (invariant) representation. To this end, we propose utilizing cross-domain consensus to learn a domain-agnostic generic prototype. Subsequently, this prototype is refined by incorporating cross-domain style variations, thereby achieving cross-view feature augmentation. Additionally, we further enhance the discriminative power of the augmented representation by formulating an identity attribute constraint to impose attention on the importance of individual attributes, while maintaining overall consistency across all pedestrians. Extensive experiments validate that the proposed CDVM model outperforms existing state-of-the-art methods by significant margins
Intelligent Performance Optimisation for Non-orthogonal Multiple Access Based IoT Networks
Get PDFThe next-generation network (NGN) is envisioned as a community of massive, data-hungry and time sensitive devices, driving the need for more efficient transmission strategies in sixth-generation (6G) wireless systems. As a key enabling technology, non-orthogonal multiple access (NOMA) allows multiple users to share the same resource block, thereby improving spectral efficiency, capacity, and connectivity. However, these benefits rely on complex optimisation under dynamic and uncertain network environments. Reinforcement learning (RL), with its ability to adapt to uncertain environments and learn efficient strategies through interaction, offers a powerful tool for addressing these challenges. This thesis focuses on integrating RL with NOMA to jointly exploit their advantages and develop intelligent transmission and resource allocation schemes for the next-generation wireless systems. First, this thesis investigates the application of deep reinforcement learning (DRL) for joint beam selection and power allocation in overload NOMA networks. A proximal policy optimisation (PPO) based solution is proposed to learn energy efficient beamforming schemes from random channel state information (CSI), achieving a 27.75% improvement in energy efficiency (EE) and consistent performance gains with increasing number of users. Second, to further enhance system efficiency, a pure-NOMA beamforming scheme is formulated by accommodating more users within the same spectrum. A curiosity-driven learning method is proposed to address the high-dimensional and non-convex optimisation problem, achieving superior performance in adapting to an EE beamforming scheme and avoiding falling into local minima. Third, the age-of-information (AoI) performance is investigated through a NOMA-assisted semi-grant-free (SGF) framework. A hierarchical learning algorithm is proposed to jointly optimize beamforming and transmission scheduling. Simulation results indicate a 31.82% improvement in overall throughput while maintaining information freshness. The outcomes of this thesis demonstrate the effectiveness of RL-based strategies in optimizing NOMA resource allocation and transmission schemes under dynamic and uncertain environments. This thesis highlights how learning-driven methods can adapt to complex resource allocation challenges and support emerging access strategies such as grant-free protocols. These findings point to the importance of future research on robust physical-layer modeling, adaptive RL designs, and efficient, reproducible, and transferable AI solutions for real-world deployment
DESIGN, SYNTHESIS AND EVALUATION OF NEW CHEMICAL TOOLS FOR NEUROBIOLOGY AND DENTISTRY APPLICATIONS
No full textThis thesis presents the development of novel chemical tools designed to access structurally diverse scaffolds with potential applications in neuroscience and dental biomaterials. Chapter one focuses on the synthesis of voltage-sensitive dyes (VSDs), particularly RH-1691 and a series of newly derived analogues. Understanding neuronal activity remains a central challenge in neuroscience due to the intricate architecture of brain networks. VSDs offer a powerful means of capturing neuronal dynamics with high temporal and spatial resolution. Building upon the seminal synthetic approach introduced by the Arseniyadis group in 2009, the first practical route to RH-1691, we have expanded this methodology to synthesize nine new VSD derivatives. Additionally, various salt forms were prepared to investigate physicochemical and biological properties. A distinguishing feature of RH-1691 is the sulfonic acid group on the pyrolozone ring, which is initially protected as a neopentylsulfonate ester and subsequently deprotected to yield the active dye. Both protected intermediates and final dyes have been submitted for biological evaluation, contributing to the development of advanced imaging tools for neuroscience research. Chapter Two focuses on the design and synthesis of novel halogenated methacrylate and acrylate monomers aimed at improving the functionality of dental resin infiltrants. A series of these halogenated monomers were synthesized and systematically evaluated for their physicochemical properties. This work addresses a key limitation of conventional resin infiltrants: their radiolucency, which hinders effective monitoring via radiographic imaging. By incorporating radiopaque monomers into resin formulations, we developed modified infiltrants that are readily detectable on X-rays. These radiopaque monomers were successfully copolymerized with standard infiltrant systems, enabling enhanced clinical monitoring and offering potential for early detection of caries progression
