As gravitational-wave catalogs grow, they will become increasingly computationally expensive to analyze in their entirety, especially when inferring astrophysical source populations with high-dimensional, flexible models. Bayesian statistics offers a natural remedy, letting us update our knowledge of physical models as new data arrive, without re-analyzing existing data. However, doing so requires the posterior probability density of model parameters for previous observations, which is typically intractable. Here, we use variational neural posterior estimation to rapidly update the inferred population of binary black holes as data are observed in gravitational-wave detectors. We apply this approach to real and simulated catalogs analyzed with both low- and high-dimensional population models, testing the reliability of three update cadences: with new catalogs of sources, month by month during an observing run, and as each new signal arrives. We investigate the success and failure modes of neural sequential updates, finding that the robustness of updating is sensitive to the information contained in each update and that updating is most effective when performed with larger segments of data. We outline one additional scientific application enabled by Bayesian updating: identification of events that are individually informative about the population. Neural Bayesian updates to astrophysical population models also provide efficient likelihood representations for joint analyses with other data, e.g., standard-siren cosmology, and similar methods can be used to perform Bayesian stochastic background searches.
BACKGROUND: Age-specific population forecasts for small areas or subnational regions are a valuable tool for local governments. However, typical population projection methods based on the cohort-component approach are difficult to apply on a smaller subnational scale. OBJECTIVE: We introduce Bayesian methods suitable for obtaining reliable age-specific population forecasts for small regions using the cohort-component method. METHODS: Our approach improves fertility forecasting by extending the Lee–Carter model with an age-region interaction term. We propose to forecast net-migration counts using skewed error terms, and introduce a Dirichlet regression to model migration age patterns as well as age proportions of fertility. RESULTS: We run our model to produce age-specific population forecasts for a set of 13 heterogeneous regions in Bavaria, Germany. We compare our method with other standard approaches and find that it produces superior out-of-sample forecasts according to both point measures and scoring rules. CONCLUSIONS: The findings suggest that the proposed Bayesian methods offer good predictive accuracy and are suitable in obtaining precise forecasts of age-specific population for smaller geo-graphical regions. CONTRIBUTION: We introduce a new method for the probabilistic projection of subnational population that works well and outperforms other current methods.
Third-generation ground-based gravitational wave detectors are expected to observe $\mathcal{O}(10^5)$ of overlapping signals per year from a multitude of astrophysical sources that will be computationally challenging to resolve individually. On the other hand, the stochastic background resulting from the entire population of sources encodes information about the underlying population, allowing for population parameter inference independent and complementary to that obtained with individually resolved events. Parameter estimation in this case is still computationally challenging, as computing the power spectrum involves sampling $\sim 10^5$ sources for each set of hyperparameters describing the binary population. In this work, we build on recently developed importance sampling techniques to compute the SGWB efficiently and train neural networks to interpolate the resulting background. We show that a multi-layer perceptron can encode the model information, allowing for significantly faster inference. We test the network assuming an observing setup with CE and ET sensitivities, where for the first time we include the intrinsic variance of the SGWB in the inference, as in this setup it presents a dominant source of measurement noise.
Hot subdwarf B stars (sdBs) are helium-burning stars with thin hydrogen-rich envelopes. Their most widely accepted formation channels involve binary evolution and progenitors near the tip of the red giant branch, thus studying these objects improves our knowledge of complicated astrophysical processes such as common envelope evolution and the helium flash. In this work, we compare the observed sdB population with a synthetic Galactic population generated through the binary population synthesis code COMPAS, which allows us to estimate the physical properties of the current-day Galactic sdB population. We show that our synthetic sdB population matches the general properties of the observations quite well in the Kiel diagram when either a normal or lognormal distribution is assumed for the assignment of hydrogen-rich envelope masses. We also find that the canonical mass assumption should only be confidently assumed for specific system configurations and that the estimated number of sdBs found within 500 pc of the Sun in our model is at least four times higher than the observational one. We recover the observational P-q relation for sdBs plus main-sequence companions, while a similar relation between sdBs and helium white dwarf companions is rather complicated. We conclude that a better understanding of hydrogen-rich envelopes is needed, as well as an observational characterization of the sdB plus main-sequence companions earlier than spectral type $~$F. These issues aside, atmospheric properties, companion types, period, and mass distributions are in good agreement with observational and theoretical studies available in the literature.
Yuki Ohnishi, Michael O. Harhay, Guangyu Tong
et al.
Recurrent events often serve as key endpoints in clinical studies but may be prematurely truncated by terminal events such as death, creating selection bias and complicating causal inference. To address this challenge, we develop a Bayesian nonparametric framework to address potential selection bias due to truncation by death within the continuous-time principal stratification framework. We introduce causal estimands for recurrent events in the presence of a terminal event and derive a partial identification result for the estimand under a dual-frailty framework, enabling transparent sensitivity analysis for non-identifiable parameters. We then propose a flexible Bayesian nonparametric prior, the enriched dependent Dirichlet process, specifically designed for joint modeling of recurrent and terminal events, addressing a limitation where standard Dirichlet process priors create random partitions dominated by recurrent events, yielding poor predictive performance for terminal events. Simulations are carried out to show that our method has superior performance compared to existing methods. We apply the proposed new Bayesian nonparametric methods to infer the causal effect of a structured exercise program on rehospitalizations, which are subject to truncation by death.
We consider the problem of privacy protection in Reinforcement Learning (RL) algorithms that operate over population processes, a practical but understudied setting that includes, for example, the control of epidemics in large populations of dynamically interacting individuals. In this setting, the RL algorithm interacts with the population over $T$ time steps by receiving population-level statistics as state and performing actions which can affect the entire population at each time step. An individual's data can be collected across multiple interactions and their privacy must be protected at all times. We clarify the Bayesian semantics of Differential Privacy (DP) in the presence of correlated data in population processes through a Pufferfish Privacy analysis. We then give a meta algorithm that can take any RL algorithm as input and make it differentially private. This is achieved by taking an approach that uses DP mechanisms to privatize the state and reward signal at each time step before the RL algorithm receives them as input. Our main theoretical result shows that the value-function approximation error when applying standard RL algorithms directly to the privatized states shrinks quickly as the population size and privacy budget increase. This highlights that reasonable privacy-utility trade-offs are possible for differentially private RL algorithms in population processes. Our theoretical findings are validated by experiments performed on a simulated epidemic control problem over large population sizes.
We reproduce the luminosity functions of the early-time peak optical luminosity, the late-time UV plateau luminosity, and the peak X-ray luminosity of tidal disruption events, using an entirely first-principles theoretical approach. We do this by first fitting three free parameters of the tidal disruption event black hole mass distribution using the observed distribution of late time UV plateau luminosities, using a time-dependent relativistic accretion model. Using this black hole mass distribution we are then, with no further free parameters of the theory, able to reproduce exactly the peak X-ray luminosity distribution of the tidal disruption event population. This proves that the X-ray luminosity of tidal disruption events are sourced from the same accretion flows which produce the late time UV plateau. Using an empirical scaling relationship between peak optical luminosities and black hole masses, itself calibrated using the same relativistic accretion theory, we are able to reproduce the observed peak optical luminosity function, again with no additional free parameters. Implications of these results include that there is no tidal disruption event "missing energy problem", that the optical and X-ray selected tidal disruption event populations are drawn from the same black hole mass distribution, that the early time optical luminosity in tidal disruption events is somewhat simple, at least on the population level, and that future LSST observations will be able to constrain the black hole mass function at low masses.
Objectives and Approach
Neighborhood influences pregnancy outcomes through effects on maternal health. We evaluated prevalence of >140 congenital anomalies monitored by the population-based Texas Birth Defects Registry among residents of Hispanic/Latino enclaves (census tracts with high proportions of Hispanic/Latino residents, immigrants, and Spanish-speaking households) and tracts with low socioeconomic status (low nSES). We included all cases regardless of pregnancy outcome (1999-2018) and a reference population of all livebirths, identified by linkage to Texas vital records. We calculated Yost socioeconomic index and Hispanic/Latino enclave index scores using linked U.S. Census Bureau data. We used Poisson regression to estimate the prevalence ratio (PR) and 95% confidence interval of anomalies among residents of: low nSES/enclave (34.5%); low nSES/non-enclave (25.6%); high nSES/enclave (4.6%); and high nSES/non-enclave (referent; 35.3%) tracts. We adjusted for maternal age, education, race/ethnicity, parity, and pre-pregnancy body mass index.
Results
Low nSES was associated with microcephaly, pulmonary artery anomalies, atrial septal defect, and Down syndrome (PRs 1.1-1.6), and inversely associated with hypospadias and other male genital anomalies (PRs 0.8-0.9). Offspring of women in low nSES enclaves were at increased risk of several congenital heart defects relative to those in other neighborhoods. Conversely, prevalence of hypospadias was lowest among offspring of women in low nSES enclaves.
Conclusions
Linkage between a population-based registry, vital records, and Census Bureau data revealed novel associations between neighborhood characteristics and congenital anomalies.
Implications
Further research is warranted to understand mechanisms linking neighborhood factors to congenital anomalies.
Vision-language (V+L) pretraining models have achieved great success in supporting multimedia applications by understanding the alignments between images and text. While existing vision-language pretraining models primarily focus on understanding objects in images or entities in text, they often ignore the alignment at the level of events and their argument structures. In this work, we propose a contrastive learning framework to enforce vision-language pretraining models to comprehend events and associated argument (participant) roles. To achieve this, we take advantage of text information extraction technologies to obtain event structural knowledge, and utilize multiple prompt functions to contrast difficult negative descriptions by manipulating event structures. We also design an event graph alignment loss based on optimal transport to capture event argument structures. In addition, we collect a large event-rich dataset (106,875 images) for pretraining, which provides a more challenging image retrieval benchmark to assess the understanding of complicated lengthy sentences. Experiments show that our zero-shot CLIP-Event outperforms the state-of-the-art supervised model in argument extraction on Multimedia Event Extraction, achieving more than 5% absolute F-score gain in event extraction, as well as significant improvements on a variety of downstream tasks under zero-shot settings.
Donghao Qiao, Amtul Haq Ayesha, Farhana Zulkernine
et al.
We propose an end-to-end framework to measure people's vital signs including Heart Rate (HR), Heart Rate Variability (HRV), Oxygen Saturation (SpO2) and Blood Pressure (BP) based on the rPPG methodology from the video of a user's face captured with a smartphone camera. We extract face landmarks with a deep learning-based neural network model in real-time. Multiple face patches also called Regions-of-Interest (RoIs) are extracted by using the predicted face landmarks. Several filters are applied to reduce the noise from the RoIs in the extracted cardiac signals called Blood Volume Pulse (BVP) signal. The measurements of HR, HRV and SpO2 are validated on two public rPPG datasets namely the TokyoTech rPPG and the Pulse Rate Detection (PURE) datasets, on which our models achieved the following Mean Absolute Errors (MAE): a) for HR, 1.73Beats-Per-Minute (bpm) and 3.95bpm respectively; b) for HRV, 18.55ms and 25.03ms respectively, and c) for SpO2, an MAE of 1.64% on the PURE dataset. We validated our end-to-end rPPG framework, ReViSe, in daily living environment, and thereby created the Video-HR dataset. Our HR estimation model achieved an MAE of 2.49bpm on this dataset. Since no publicly available rPPG datasets existed for BP measurement with face videos, we used a dataset with signals from fingertip sensor to train our deep learning-based BP estimation model and also created our own video dataset, Video-BP. On our Video-BP dataset, our BP estimation model achieved an MAE of 6.7mmHg for Systolic Blood Pressure (SBP), and an MAE of 9.6mmHg for Diastolic Blood Pressure (DBP). ReViSe framework has been validated on datasets with videos recorded in daily living environment as opposed to less noisy laboratory environment as reported by most state-of-the-art techniques.
Population protocols are a relatively novel computational model in which very resource-limited anonymous agents interact in pairs with the goal of computing predicates. We consider the probabilistic version of this model, which naturally allows to consider the setup in which a small probability of an incorrect output is tolerated. The main focus of this thesis is the question of confident leader election, which is an extension of the regular leader election problem with an extra requirement for the eventual leader to detect its uniqueness. Having a confident leader allows the population protocols to determine the convergence of its computations. This behaviour of the model is highly beneficial and was shown to be feasible when the original model is extended in various ways. We show that it takes a linear in terms of the population size number of interactions for a probabilistic population protocol to have a non-zero fraction of agents in all reachable states, starting from a configuration with all agents in the same state. This leads us to a conclusion that confident leader election is out of reach even with the probabilistic version of the model.
Isabel Youngs, Christopher Dick, Ronald Prevost
et al.
The importance of the decennial census is clear, but the 2020 Census faced unprecedented challenges with unknown effects on data quality. To assist data users in identifying deviations between expected counts and the released counts across population and housing indicators, we developed the 2020 Census County Assessment Tool. The tool offers contextual data for each county in the US on factors which could have contributed to census collection issues, including COVID-19, internet-first response, natural disasters, and broadband access rates. The tool also allows users to quickly see the distributions of divergence across race, ethnicity, and housing, allowing a user to dive into the data for their specific use. The tool compiles this information into downloadable reports and points users to additional local data sources or experts to seek more assistance.
This tool assists planners, administrators, policymakers, teachers and students, reporters, researchers, and anyone interested in census data in identifying strategies for utilizing census data appropriately. This tool particularly provides resources to data users interested in the accurate enumeration of race and ethnicity, pointing users to areas which may need additional attention for the 2030 census. This tool can assist users in intercensal planning, count review requests, population estimates challenges, funding allocation corrections, Get Out the Count investment strategies, and more. This presentation will discuss the development process and methodology of the tool, as well as our plans for future additions and other tools we are in the process of creating.
It was not officially possible to leave the cadastral territory for recreational purposes in Czechia during the period from the 1 March to the 21 March 2021. The aim of this study was to evaluate how this lockdown affected the amount of time young people spent outdoors and their health and mental well-being. Our research was aimed at students at all levels of school. Immediately after the end of the strictest phase of the lockdown, we conducted a questionnaire survey and collected data from more than a thousand students at elementary schools, secondary schools and universities, as well as 160 parents of 269 pre-school and primary school children. The answers to the close-ended questions were evaluated by statistical analysis, while the answers to the open-ended questions were evaluated using thematic analysis. The results show that the impact of restrictive measures on the health and psyche of young people was significant, especially for female students. Lockdown significantly reduced respondents’ opportunities to spend time outdoors. Male students spent significantly more of their free time in front of computer screens. Respondents living in buildings without a garden and young people who could not use a recreational building outside the district of residence were most affected by restrictions during the lockdown.
Jacyra Araujo, Elisângela da Silva Rodrigues, Luis Fernando Silva Castro-de-Araujo
et al.
Background
To investigate the clinical epidemiological characteristics of a large data set of visits to outpatient children mental health services in Brazil, as well as to identify relevant relationships between age, sex and three common mental disorders in childhood: pervasive developmental disorders, ADHD and mild depressive disorders.
Methods
We extracted data from a public repository, DATASUS, regarding child outpatient mental health services in Brazil, from 2008 to 2012. We performed an analysis of the number of visits per inhabitant and inferential analyses with logistic regressions for ADHD (F90.0), Pervasive Developmental Disorders (F84.0-F84.9), and Mild Depressive Episode (F32.0) as outcomes, controlling for age, year of the visit, number of new CAPSI stratified by region.
Findings
Attention-deficit hyperactivity disorder (ADHD) was the most common condition identified across the country. The analyses by region showed a high number of visits due to mental retardation in the Northeast and depressive episodes in the South. Regressions showed that older children are less likely to visit outpatient services with a diagnosis of ADHD (F90.0).
Conclusions
Our analysis shows the conditions which cause the most burden to the child psychiatry outpatient centers in Brazil and relevant differences between regions. This information has immediate use for the training of staff and allocation of resources in each region.
For collaborations please contact: Email: jacyra.paiva@fiocruz.brFor collaborations please contact: Email: jacyra.paiva@fiocruz.br
The literature shows outstanding capabilities for CNNs in event recognition in images. However, fewer attempts are made to analyze the potential causes behind the decisions of the models and exploring whether the predictions are based on event-salient objects or regions? To explore this important aspect of event recognition, in this work, we propose an explainable event recognition framework relying on Grad-CAM and an Xception architecture-based CNN model. Experiments are conducted on three large-scale datasets covering a diversified set of natural disasters, social, and sports events. Overall, the model showed outstanding generalization capabilities obtaining overall F1-scores of 0.91, 0.94, and 0.97 on natural disasters, social, and sports events, respectively. Moreover, for subjective analysis of activation maps generated through Grad-CAM for the predicted samples of the model, a crowdsourcing study is conducted to analyze whether the model's predictions are based on event-related objects/regions or not? The results of the study indicate that 78%, 84%, and 78% of the model decisions on natural disasters, sports, and social events datasets, respectively, are based onevent-related objects or regions.
María Eugenia Fernández, Jorge Osvaldo Gentili, Ana Casado
et al.
The objective of this work is to analyze the spatio-temporal distribution of Global Horizontal Irradiation (GHI) on a regional scale and its relationship with frequent synoptic situations in the south of the Pampeana region (Argentina). It was verified that the latitudinal pattern of distribution of the GHI is modified in the region by cloud cover, which is in turn determined by the seasonal dynamics of action centers and the passage of fronts in summer and winter. The South America Monsoon System (SAMS) defines differential situations of cloudiness and rainfall in the region, which affect GHI. GHI increased successively between the decades 1981–2010, a factor associated with the variability of rainfall that characterizes the region.
Knowing the state of our health at every moment in time is critical for advances in health science. Using data obtained outside an episodic clinical setting is the first step towards building a continuous health estimation system. In this paper, we explore a system that allows users to combine events and data streams from different sources to retrieve complex biological events, such as cardiovascular volume overload. These complex events, which have been explored in biomedical literature and which we call interface events, have a direct causal impact on relevant biological systems. They are the interface through which the lifestyle events influence our health. We retrieve the interface events from existing events and data streams by encoding domain knowledge using an event operator language.
The large deviation principle on phase space is proved for a class of Markov processes known as random population dynamics with catastrophes. In the paper we study the process which corresponds to the random population dynamics with linear growth and uniform catastrophes, where an eliminating portion of the population is chosen uniformly. The large deviation result provides an optimal trajectory of large fluctuation: it shows how the large fluctuations occur for this class of processes.
Educated and talented people drive progress in every country. That’s why no
country can neglect these people; that would mean losing one’s own potential.
This paper emphasises the importance of educating scholarship students abroad
as a means of developing and accumulating human resources and a key
determinant of sustainable development in the modern world. Investing in the
education of the best students (scholarship holders) is an investment in the
future, which brings multiple benefits on a social, economic, and political
level. Migration is an important phenomenon that attracts public attention,
especially when it comes to highly educated experts leaving their home
country in search of better education. Highly educated experts have been
leaving Serbia for several decades, which poses an obvious problem for local
society. The topic of permanent migration is dominant both in foreign and
domestic literature, but studies on the temporary migration of highly
educated students (scholarship holders) is almost nonexistent. The aim of
this paper is to point out the value of returnee scholarship holders and the
importance of creating the appropriate conditions for them to stay in the
country. A survey conducted on a group of 96 returnee scholarship holders
identified factors that affect their satisfaction with living in Serbia. The
survey involved experts from Serbia who were educated abroad as scholarship
holders, where they acquired academic titles and are now employed: as faculty
teachers (32); as researchers at scientific institutes (24); in the private
sector (21); at universities (12); in state administrative departments (5);
and in medical institutions (2). The criterion for selecting this group of
respondents was that they had stayed abroad as scholarship holders, whether
they used scholarships from domestic (24) or foreign (72) funds. Scholarship
students go abroad mostly because of their personal aspirations for training,
gaining new experiences, and because of the inability to study the desired
discipline in their country, as was the case for 74 respondents. The main
reasons for deciding to return are family (25) and the belief that they have
a good chance to work in Serbia (18), while 16 respondents could not stay
abroad. In this paper, we used the factor analysis method. The main factors
that create satisfaction with life in Serbia are isolated. These factors are:
satisfaction with work and a set of factors that strongly correlate with it
(the ability to make decisions, the implementation of acquired knowledge,
peer acceptance), as well as the recognition of their diploma in Serbia
without any difficulties. By improving these factors, there might be a
significant increase in the chance that returnee scholarship holders remain
in Serbia for a long time. Based on this, it would be wise to build a
strategy on how to encourage returnee scholarship holders to stay in the
country. The results obtained in this study represent a contribution to a
search for a strategy that will attract, involve, and retain educated people
in the country. [Project of the Serbian Ministry of Education, Science and
Technological Development, Grant no. TR 35030]