143174 research outputs found
Sort by
TeleSparse: Practical Privacy-Preserving Verification of Deep Neural Networks
No full textVerification of the integrity of deep learning inference is crucial for understanding whether a model is being applied correctly. However, such verification typically requires access to model weights and (potentially sensitive or private) training data. So-called Zeroknowledge Succinct Non-Interactive Arguments of Knowledge (ZK-SNARKs) would appear to provide the capability to verify model inference without access to such sensitive data. However, applying ZK-SNARKs to modern neural networks, such as transformers and large vision models, introduces significant computational overhead. We present TeleSparse, a ZK-friendly post-processing mechanisms to produce practical solutions to this problem. TeleSparse tackles two fundamental challenges inherent in applying ZK-SNARKs to modern neural networks: (1) Reducing circuit constraints: Overparameterized models result in numerous constraints for ZK-SNARK verification, driving up memory and proof generation costs. We address this by applying sparsification to neural network models, enhancing proof efficiency without compromising accuracy or security. (2) Minimizing the size of lookup tables required for non-linear functions, by optimizing activation ranges through neural teleportation, a novel adaptation for narrowing activation functions' range. TeleSparse reduces prover memory usage by 67% and proof generation time by 46% on the same model, with an accuracy trade-off of approximately 1%. We implement our framework using the Halo2 proving system and demonstrate its effectiveness across multiple architectures (Vision-transformer, ResNet, MobileNet) and datasets (ImageNet,CIFAR-10,CIFAR-100). This work opens new directions for ZK-friendly model design, moving toward scalable, resourceefficient verifiable deep learning
Current status of primary, secondary and tertiary prevention of congenital cytomegalovirus disease: a call to action
No full textPurpose of review
Globally, sequelae of congenital CMV (CCMV) impact an estimated 350 000 children born annually. In this review, we consider new evidence across primary, secondary and tertiary prevention approaches, and remaining evidence gaps.
Recent findings
Education on hygiene precautions can reduce risk of primary CMV acquisition in pregnancy, and may have a role in some settings in reducing CCMV cases resulting from nonprimary infection, but public and health worker knowledge and awareness remains low. Evidence that valaciclovir treatment can reduce CMV vertical transmission has led to renewed interest in antenatal CMV screening in some high-income countries over recent years, although there is a lack of recommendation in most guidelines and significant evidence gaps remain. Newborn CCMV screening has been adopted in some states/provinces in Canada/USA, with first results recently published. Newborn prognostic scoring systems are evolving, with potential for more effective targeting of newborn treatment and tertiary prevention of CCMV disease.
Summary
We make suggestions for clinical practice and research, particularly to address evidence gaps around: safety and effectiveness of antenatal CMV screening and antiviral prophylaxis; findings relating to detection of nonprimary infection in pregnancy; new prognostic neonatal scoring systems; and learning from follow-up of children born into state-wide universal CMV screening programmes
Tree Stem Methane Flux Dynamics Across Functional Types and Seasons in a Reforested Upland Atlantic Forest, Brazil
No full textImperial Users onl
Unpacking the predictors of loneliness: an inferential analysis from the INTERACT study
No full textBackground:
Loneliness is a pressing public health concern with wide-ranging impacts on mental, physical and social
wellbeing. Building on the INTERACT Study-the largest UK-based investigation of loneliness-this paper
explores demographic, social and health-related predictors of loneliness and social capital, using
multiple validated measures.
Methods:
We analysed cross-sectional data from 135,722 community-dwelling adults across England. Loneliness
was assessed using both the UCLA 3-item Loneliness Scale and the ONS Direct Measure of Loneliness
(DMOL). Social capital was measured using a composite scale of neighbourhood trust, cohesion and
reciprocity. Multivariable ordinal logistic regression was used to examine predictors of loneliness; binary
logistic regression was used to analyse correlates of high versus low social capital.
Results:
Younger age (particularly 16–25), being single, unemployed or living with disability were consistently
associated with higher loneliness across both scales. In contrast, greater social contact having nine or
more friends or relatives was strongly protective (UCLA: aOR 0.09; DMOL: aOR 0.16). University
education was associated with higher loneliness on the UCLA scale but lower loneliness on the DMOL.
High social capital was more prevalent among older, married and retired individuals and strongly
predicted lower loneliness. Respondents with long-term conditions or disability had reduced odds of
high social capital (aORs 0.65 and 0.59 respectively).
Conclusions:
This study highlights consistent sociodemographic and social predictors of loneliness, as well as the
protective role of social capital. Findings support the need for targeted public health interventions that
address social connection among young adults, single people, the unemployed and individuals in poor
health. Strategies that invest in neighbourhood cohesion and social infrastructure are vital for mitigating
loneliness and strengthening community wellbeing
Intergenerational mobility in 28 European countries – a new measure from a public health perspective
No full textIntroduction:
We assessed levels of intergenerational mobility across 28 European countries using a new composite measure of social status for European adults and their parents.
Methods:
We conducted a secondary dataset analysis of the cross-sectional Special Eurobarometer Survey wave 88.4, performed in December 2017 among 27538 individuals in 28 countries. Respondents’ adulthood perceived social status (‘adult’) as well as their parents’ perceived social status (‘parental’) were investigated by combining three indicators: subjective social status, occupation, and education. We created an intergenerational mobility scale by calculating the change in respondents’ social status compared to their parents’. The size of the interquartile range (IQR) of this scale was used to estimate intergenerational mobility in each country.
Results:
Median parental and adult social status scores varied across countries and regions. Overall, the median adult and parental social status scores were 4.35 (out of 10) and 5.00. Denmark (3.15), the Netherlands (2.96), and Sweden (2.96) had the largest IQRs of intergenerational mobility scores, indicating greater intergenerational mobility, whereas Slovakia (1.85), Croatia (1.85), Hungary (1.94) and Bulgaria (1.94) had the smallest IQRs.
Conclusions:
Using a new composite indicator of intergenerational mobility, we found large variation in intergenerational mobility between countries, as expressed by the size of the IQR of our novel scale. Our findings highlight countries which could reorient their policies to increase social mobility
A stochastic analysis approach to tensor field theories
No full textWe present two different arguments using stochastic analysis to construct super-renormalizable tensor field theories, namely the T4/3 and T4/4 models. The first approach is the construction of a Langevin dynamic [15 , 23 ] combined with a PDE energy estimate while the second is an application of the variational approach of Barashkov and Gubinelli [ 3]. By leveraging the melonic structure of divergences, regularizing properties of non-local products, and controlling certain random operators, we demonstrate that for tensor field theories these approaches can be significantly simplified in comparison to what is required for Φ4/d models
Impact of dhps mutations on sulfadoxine pyrimethamine protective efficacy and implications for malaria chemoprevention
No full textSulfadoxine-pyrimethamine (SP) is recommended for perennial malaria chemoprevention in young children in high burden areas across Africa. Mutations in the dihydropteroate synthase (dhps) gene (437 G/540E/581 G) associated with sulfadoxine resistance vary regionally, but their effect on SP protective efficacy is unclear. We retrospectively analyse time to microscopy and PCR-confirmed re-infection in seven efficacy trials including 1639 participants in 12 sites across Africa. We estimate the duration of SP protection against parasites with different genotypes using a Bayesian mathematical model that accounts for variation in transmission intensity and genotype frequencies. The longest duration of SP protection is >42 days against dhps sulfadoxine-susceptible parasites and 30.3 days (95%Credible Interval (CrI):17.1-45.1) against the West-African genotype dhps GKA (437G-K540-A581). A shorter duration of protection is estimated against parasites with additional mutations in the dhps gene, with 16.5 days (95%CrI:11.2-37.4) protection against parasites with the east-African genotype dhps GEA (437G-540E-A581) and 11.7 days (95%CrI:8.0-21.9) against highly resistant parasites carrying the dhps GEG (437G-540E−581G) genotype. Using these estimates and modelled genotype frequencies we map SP protection across Africa. This approach and our estimated parameters can be directly applied to any setting using local genomic surveillance data to inform decision-making on where to scale-up SP-based chemoprevention or consider alternatives
Sensitivity and optimisation of the acoustic response of short circular holes with turbulent bias flow
No full textShort circular holes with turbulent bias flows passing through them can absorb or generate acoustic energy. This property is relevant for many industrial applications, such as liners and injectors. A recent study suggested that the acoustic response of such perforations could be strongly sensitive to small modifications of the geometries of their lips. In this work, we study this sensitivity and exploit it to design holes with optimal acoustic properties. To this end, we use a numerical approach based on a two-step method, where a steady mean flow is first calculated as the solution of the incompressible RANS equations. Small-amplitude acoustic perturbations are then superimposed on this mean flow, and their dynamics are obtained by solving the linearised Navier–Stokes equations. To validate the approach, the results are compared with experiments for a hole with sharp edges, finding an excellent agreement. Subsequently, we simulate four holes with modified corners consisting in chamfers or circular fillets either of the upstream or downstream corners. The acoustic absorption is found to be strongly sensitive to all modifications with the upstream chamfer presenting the strongest variations. Finally, we combine the simulations with a Bayesian optimisation algorithm to obtain the size of the upstream chamfer that maximises the acoustic absorption. The optimised hole goes from exhibiting strong whistling to one which strongly damps acoustic energy
Assessing a mixed hydrogen economy for resilient net-zero using energy system modelling
No full textLow-carbon hydrogen will play a key role in the portfolio of alternative fuels towards climate targets. However, there are gaps in determining how deploying different hydrogen colours and production pathways contributes to achieving climate goals in a jurisdiction, including the economic impact on the energy system, the wider system, and the impact on the final energy mix. This work is the first to perform a detailed analysis of the interrelation between blue hydrogen, green hydrogen (onshore and offshore produced) and other hydrogen colours using an energy system model. Results show that for scenarios where blue hydrogen and green hydrogen are not deployed, green hydrogen is not deployed, blue hydrogen is not deployed, and blue hydrogen, onshore and offshore green hydrogen are deployed, have energy system costs 0.94 %, 0.32 %, 0.06 % and −0.1 % respectively from a reference with the base hydrogen types, blue and only onshore green available. Our optimum 'Mixed Hydrogen Economy' scenario has lowest decarbonisation costs by 2050 when net-zero emissions is achieved and ensures diversity of energy supply to boost energy security. Results emphasise the hydrogen economy is most pronounced when all available and sustainable hydrogen production options are deployed. The findings contribute towards evaluating the impact of hydrogen deployment in low-emissions scenarios
Framework for understanding public pluralities in greenspace design and consultation
No full textPublic participation underpins contemporary governance, facilitating discourse, enhancing decision legitimacy, and improving outcomes. In urban greenspace planning, however, persistent barriers, including cultural disparities and tokenism, impede effective engagement. Traditional assumptions of a homogeneous ‘public’ and uncritical endorsement of participation’s merits overlook group heterogeneity. Acknowledging public plurality is essential for authentic involvement. Within environmental science, scant research addresses public segmentation to devise bespoke, inclusive strategies. This study innovatively integrates Situational Theory of Public and Self-Determination Theory to examine diverse public segments in green space consultation and design. A UK-based survey classified four segments (active, aware, latent, non-public) and linked them to motivational drivers: trust, recognition, cohesion, employing ANOVA and PCA. Results reveal uniform approaches are inadequate. Autonomy-supportive contexts and community bonds outweigh generic incentives. Extrinsic rewards prove secondary to intrinsic factors. This framework offers practitioners a tool to tailor engagement, optimising greenspace outcomes across diverse populations