ABSTRACT
In the field of autofocus for optical systems, although passive focusing methods are widely used due to their cost-effectiveness, fixed focusing windows and evaluation functions in certain scenarios can still lead to focusing failures. Additionally, the lack of datasets limits the extensive research of deep learning methods. In this work, we propose a neural network autofocus method with the capability of dynamically selecting the region of interest (ROI). Our main work is as follows: first, we construct a dataset for automatic focusing of grayscale images; second, we transform the autofocus issue into an ordinal regression problem and propose two focusing strategies: full-stack search and single-frame prediction; and third, we construct a MobileViT network with a linear self-attention mechanism to achieve automatic focusing on dynamic regions of interest. The effectiveness of the proposed focusing method is verified through experiments, and the results show that the focusing MAE of the full-stack search can be as low as 0.094, with a focusing time of 27.8 ms, and the focusing MAE of the single-frame prediction can be as low as 0.142, with a focusing time of 27.5 ms.
ABSTRACT
AbstractMorphology often reflects ecology, enabling the prediction of ecological roles for taxa that lack direct observations, such as fossils. In comparative analyses, ecological traits, like diet, are often treated as categorical, which may aid prediction and simplify analyses but ignores the multivariate nature of ecological niches. Furthermore, methods for quantifying and predicting multivariate ecology remain rare. Here, we ranked the relative importance of 13 food items for a sample of 88 extant carnivoran mammals and then used Bayesian multilevel modeling to assess whether those rankings could be predicted from dental morphology and body size. Traditional diet categories fail to capture the true multivariate nature of carnivoran diets, but Bayesian regression models derived from living taxa have good predictive accuracy for importance ranks. Using our models to predict the importance of individual food items, the multivariate dietary niche, and the nearest extant analogs for a set of data-deficient extant and extinct carnivoran species confirms long-standing ideas for some taxa but yields new insights into the fundamental dietary niches of others. Our approach provides a promising alternative to traditional dietary classifications. Importantly, this approach need not be limited to diet but serves as a general framework for predicting multivariate ecology from phenotypic traits.
Subject(s)
Ecosystem , Mammals , Animals , Bayes Theorem , Diet , Food , Fossils , Phylogeny , EcologyABSTRACT
Many high-throughput genomic applications involve a large set of potential covariates and a response which is frequently measured on an ordinal scale, and it is crucial to identify which variables are truly associated with the response. Effectively controlling the false discovery rate (FDR) without sacrificing power has been a major challenge in variable selection research. This study reviews two existing variable selection frameworks, model-X knockoffs and a modified version of reference distribution variable selection (RDVS), both of which utilize artificial variables as benchmarks for decision making. Model-X knockoffs constructs a 'knockoff' variable for each covariate to mimic the covariance structure, while RDVS generates only one null variable and forms a reference distribution by performing multiple runs of model fitting. Herein, we describe how different importance measures for ordinal responses can be constructed that fit into these two selection frameworks, using either penalized regression or machine learning techniques. We compared these measures in terms of the FDR and power using simulated data. Moreover, we applied these two frameworks to high-throughput methylation data for identifying features associated with the progression from normal liver tissue to hepatocellular carcinoma to further compare and contrast their performances.
Subject(s)
Biomarkers, Tumor/standards , High-Throughput Screening Assays/standards , Animals , Data Interpretation, Statistical , False Positive Reactions , High-Throughput Screening Assays/methods , Humans , Machine LearningABSTRACT
Continuous response data are regularly transformed to meet regression modeling assumptions. However, approaches taken to identify the appropriate transformation can be ad hoc and can increase model uncertainty. Further, the resulting transformations often vary across studies leading to difficulties with synthesizing and interpreting results. When a continuous response variable is measured repeatedly within individuals or when continuous responses arise from clusters, analyses have the additional challenge caused by within-individual or within-cluster correlations. We extend a widely used ordinal regression model, the cumulative probability model (CPM), to fit clustered, continuous response data using generalized estimating equations for ordinal responses. With the proposed approach, estimates of marginal model parameters, cumulative distribution functions , expectations, and quantiles conditional on covariates can be obtained without pretransformation of the response data. While computational challenges arise with large numbers of distinct values of the continuous response variable, we propose feasible and computationally efficient approaches to fit CPMs under commonly used working correlation structures. We study finite sample operating characteristics of the estimators via simulation and illustrate their implementation with two data examples. One studies predictors of CD4:CD8 ratios in a cohort living with HIV, and the other investigates the association of a single nucleotide polymorphism and lung function decline in a cohort with early chronic obstructive pulmonary disease.
Subject(s)
Models, Statistical , Humans , Computer Simulation , Probability , UncertaintyABSTRACT
PURPOSE: Socioeconomic factors are known to modulate health. Concerning sleep apnea, influences of income, education, work, and living in a partnership are established. However, results differ between national and ethnic groups. Results also differ between various clinical studies and population-based approaches. The goal of our study was to determine if such factors can be verified in the population of Pomerania, Germany. METHODS: A subgroup from the participants of the population-based Study of Health in Pomerania volunteered for an overnight polysomnography. Their data were subjected to an ordinal regressions analysis with age, sex, body mass index (BMI), income, education, work, and life partner as predictors for the apnea-hypopnea index. RESULTS: Among the subgroup (N = 1209) from the population-based study (N = 4420), significant effects were found for age, sex, and BMI. There were no significant effects for any of the socioeconomic factors. CONCLUSION: Significant effects for well-established factors as age, sex, and BMI show that our study design has sufficient power to verify meaningful associations with sleep apnea. The lack of significant effects for the socioeconomic factors suggests their clinical irrelevance in the tested population.
Subject(s)
Sleep Apnea Syndromes , Sleep Apnea, Obstructive , Humans , Sleep Apnea, Obstructive/diagnosis , Sleep Apnea, Obstructive/epidemiology , Sleep Apnea, Obstructive/complications , Sleep Apnea Syndromes/diagnosis , Sleep Apnea Syndromes/epidemiology , Sleep Apnea Syndromes/complications , Socioeconomic Factors , Polysomnography/methods , Germany , Body Mass IndexABSTRACT
In many medical applications, interpretable models with high prediction performance are sought. Often, those models are required to handle semistructured data like tabular and image data. We show how to apply deep transformation models (DTMs) for distributional regression that fulfill these requirements. DTMs allow the data analyst to specify (deep) neural networks for different input modalities making them applicable to various research questions. Like statistical models, DTMs can provide interpretable effect estimates while achieving the state-of-the-art prediction performance of deep neural networks. In addition, the construction of ensembles of DTMs that retain model structure and interpretability allows quantifying epistemic and aleatoric uncertainty. In this study, we compare several DTMs, including baseline-adjusted models, trained on a semistructured data set of 407 stroke patients with the aim to predict ordinal functional outcome three months after stroke. We follow statistical principles of model-building to achieve an adequate trade-off between interpretability and flexibility while assessing the relative importance of the involved data modalities. We evaluate the models for an ordinal and dichotomized version of the outcome as used in clinical practice. We show that both tabular clinical and brain imaging data are useful for functional outcome prediction, whereas models based on tabular data only outperform those based on imaging data only. There is no substantial evidence for improved prediction when combining both data modalities. Overall, we highlight that DTMs provide a powerful, interpretable approach to analyzing semistructured data and that they have the potential to support clinical decision-making.
Subject(s)
Ischemic Stroke , Stroke , Humans , Neural Networks, Computer , PrognosisABSTRACT
The assessment of bone age is important for evaluating child development, optimizing the treatment for endocrine diseases, etc. And the well-known Tanner-Whitehouse (TW) clinical method improves the quantitative description of skeletal development based on setting up a series of distinguishable stages for each bone individually. However, the assessment is affected by rater variability, which makes the assessment result not reliable enough in clinical practice. The main goal of this work is to achieve a reliable and accurate skeletal maturity determination by proposing an automated bone age assessment method called PEARLS, which is based on the TW3-RUS system (analysis of the radius, ulna, phalanges, and metacarpal bones). The proposed method comprises the point estimation of anchor (PEA) module for accurately localizing specific bones, the ranking learning (RL) module for producing a continuous stage representation of each bone by encoding the ordinal relationship between stage labels into the learning process, and the scoring (S) module for outputting the bone age directly based on two standard transform curves. The development of each module in PEARLS is based on different datasets. Finally, corresponding results are presented to evaluate the system performance in localizing specific bones, determining the skeletal maturity stage, and assessing the bone age. The mean average precision of point estimation is 86.29%, the average stage determination precision is 97.33% overall bones, and the average bone age assessment accuracy is 96.8% within 1 year for the female and male cohorts.
Subject(s)
Age Determination by Skeleton , Radius , Child , Humans , Male , Female , Age Determination by Skeleton/methods , Radius/diagnostic imaging , Ulna/diagnostic imaging , Reference ValuesABSTRACT
Background: Although fruits and vegetables are considered a pillar of healthy eating, previous evidence suggests that their consumption in Eastern European countries is low, and their association with health outcomes has rarely been researched in this region. Aim: To examine the effect of fruit and vegetable intake on self-rated health (SRH) in the Czech arm of the Health, Alcohol and Psychosocial factors in Eastern Europe prospective cohort study. Methods: Dietary data on fruit and vegetable intake was measured at baseline using food frequency questionnaires, and SRH from the second wave was chosen as the main outcome. The relationship between fruit and vegetable intake and SRH was analysed using multivariable ordinal regression. Results: A total of 4255 persons aged 45-69, in good and very good SRH at baseline were included in the longitudinal analysis, with a median follow-up time of 3.7 years. In the second wave, 218 (5.1%) individuals reported poor or very poor SRH. In the fully adjusted model, individuals in the lowest fruit and vegetable intake quartile had higher odds of poor SRH compared to those in the highest quartile (OR = 1.24, 95% CI: 1.01-1.52). When examined separately, the results were similar: for vegetables (OR = 1.25, 95% CI: 1.03-1.51) and fruit (OR = 1.18, 95% CI: 0.97-1.44). Conclusion: The observed longitudinal association suggests that low fruit and vegetable intake is associated with poor SRH in the Czech Republic. Considering almost half of our sample reported less than the daily recommended intake of 400 grams of fruits and vegetables, higher consumption should be supported.
Subject(s)
Fruit , Vegetables , Humans , Prospective Studies , Czech Republic , DietABSTRACT
BACKGROUND: Oral health knowledge forms part of oral health literacy that enables individuals to inform appropriate oral health decisions and actions. Oral health-related quality of life (OHRQoL) characterizes self-perception of well-being influenced by oral health. This study aimed to examine the relationship between oral health knowledge and OHRQoL. METHODS: A random sample of 19-to-24-year-old first-year undergraduate students (n = 372) in Minnesota, United States of America was used. Each student was assessed with an online survey using the Comprehensive Measure of Oral Health Knowledge (CMOHK) and the OHRQoL items of the World Health Organization (WHO) Oral Health Questionnaire for Adults. Relationships between OHRQoL parameters and CMOHK together with other covariates were assessed using ordinal regression models. Associations between OHRQoL parameters were examined with the Kendall's tau-b method. RESULTS: Dry mouth (45%) was the most reported OHRQoL issue. The respondents showing good oral health knowledge were less likely to experience speech or pronunciation difficulty (ß=-1.12, p = 0.0006), interrupted sleep (ß=-1.43, p = 0.0040), taking days off (ß=-1.71, p = 0.0054), difficulty doing usual activities (ß=-2.37, p = 0.0002), or reduced participation in social activities due to dental or oral issues (ß=-1.65, p = 0.0078). CONCLUSIONS: This study suggested a protective effect of better oral health knowledge on specific OHRQoL issues. In addition to provision of affordable dental services, university-wide oral health education can be implemented to improve OHRQoL in undergraduate students.
Subject(s)
Oral Health , Quality of Life , Adult , Humans , United States , Young Adult , Universities , Surveys and Questionnaires , StudentsABSTRACT
Motivated by data measuring progression of leishmaniosis in a cohort of US dogs, we develop a Bayesian longitudinal model with autoregressive errors to jointly analyze ordinal and continuous outcomes. Multivariate methods can borrow strength across responses and may produce improved longitudinal forecasts of disease progression over univariate methods. We explore the performance of our proposed model under simulation, and demonstrate that it has improved prediction accuracy over traditional Bayesian hierarchical models. We further identify an appropriate model selection criterion. We show that our method holds promise for use in the clinical setting, particularly when ordinal outcomes are measured alongside other variables types that may aid clinical decision making. This approach is particularly applicable when multiple, imperfect measures of disease progression are available.
ABSTRACT
In recent years, the increasing number of natural disasters has raised concerns about the sustainability of our planet's future. As young people comprise the generation that will suffer from the negative effects of climate change, they have become involved in a new climate activism that is also gaining interest in the public debate thanks to the Fridays for Future (FFF) movement. This paper analyses the results of a survey of 1,138 young people in a southern Italian region to explore their perceptions of the extent of environmental problems and their participation in protests of green movements such as the FFF. The statistical analyses perform an ordinal classification tree using an original impurity measure considering both the ordinal nature of the response variable and the heterogeneity of its ordered categories. The results show that respondents are concerned about the threat of climate change and participate in the FFF to claim their right to a healthier planet and encourage people to adopt environmentally friendly practices in their lifestyles. Young people feel they are global citizens, connected through the Internet and social media, and show greater sensitivity to the planet's environmental problems, so they are willing to take effective action to demand sustainable policies from decision-makers. When planning public policies that will affect future generations, it is important for policymakers to know the demands and opinions of key stakeholders, especially young people, in order to plan the most appropriate measures, such as climate change mitigation.
ABSTRACT
Probabilistic loss assessments from natural hazards require the quantification of structural vulnerability. Building damage data can be used to estimate fragility curves to obtain realistic descriptions of the relationship between a hazard intensity measure and the probability of exceeding certain damage grades. Fragility curves based on the lognormal cumulative distribution function are popular because of their empirical performance as well as theoretical properties. When we are interested in estimating exceedance probabilities for multiple damage grades, these are usually derived per damage grade via separate probit regressions. However, they can also be obtained simultaneously through an ordinal model which treats the damage grades as ordered and related instead of nominal and distinct. When we use nominal models, a collapse fragility curve is constructed by treating data of "near-collapse" and "no damage" the same: as data of noncollapse. This leads to a loss of information. Using synthetic data as well as real-life data from the 2015 Nepal earthquake, we provide one of the first formal demonstrations of multiple advantages of the ordinal model over the nominal approach. We show that modeling the ordering of damage grades explicitly through an ordinal model leads to higher sensitivity to the data, parsimony and a lower risk of overfitting, noncrossing fragility curves, and lower associated uncertainty.
Subject(s)
Probability , UncertaintyABSTRACT
This paper proposes a novel end-to-end pipeline that uses the ordinal information and relative relation of images for visibility estimation (VISOR-NET). By encoding ordinal information into a set of relatively ordered image pairs, VISOR-NET can learn a global ranking function effectively. Due to the lack of real scenes or continuous labels in public foggy datasets, we collect a large-scale dataset that we term Foggy Highway Visibility Images (FHVI), which are taken from real surveillance scenes, and synthesize an INDoor Foggy images dataset (INDF) with continuous annotation. This work measures the estimation effectiveness on two public datasets and our FHVI dataset as a classification task and then on the INDF dataset as a regression task. Comprehensive experiments with existing deep-learning methods demonstrate the performance of the proposed method in terms of estimation accuracy, the convergence rate, model stability, and data requirements. Moreover, this method can extend inter-level visibility estimation to intra-level visibility estimation and can realize approximate regression estimation under discrete-level labels.
Subject(s)
Deep LearningABSTRACT
BACKGROUND: Anaemia is a serious global public health problem with high prevalence (>40%) in children particularly in low- and middle-income countries including Namibia with a current 46.1% prevalence rate. AIM: This study was aimed at examining the sociodemographic factors influencing the occurrence of childhood anaemia levels in Namibia. METHOD: A multivariate ordinal regression model was applied to statistically identify potential sociodemographic factors associated with anaemia levels among children under-5 years old using data collected from the 2013 NDHS. RESULTS: The odds of having mild anaemia level was lower for sociodemographic characteristics such as mother's age, total children ever born, health insurance coverage, child's residency, child's age and main language spoken at home. The odds of having moderate anaemia level was higher for characteristics such as mother's age, place of residence, highest education level and child's diarrhoea status, while it was lower for characteristics such as age of head of household, total children ever born, health insurance coverage and sex of child. Similarly, the odds of having severe anaemia level was higher for characteristics such as region, place of residence, highest education level, number of household members, wealth index, health insurance coverage, child's residency and child's diarrhoea status, while it was lower for characteristics such as total children ever born and sex of child. CONCLUSION: It is therefore recommended that the policies and practices concerning anaemia diagnosis, treatment and prevention in the country be substantially revised by policy-makers, starting with the known prevalent causes and identified sociodemographic factors from this study.
ABSTRACT
The COVID-19 crisis has forced many people to work from home, rather than at their regular workplace. This paper aims to assess the impact of the pandemic on telecommuting and commuting behavior after the end of the crisis: Will people embrace teleworking and reduce commuting, even to some extent, or will they resume their pre-pandemic work patterns? This study, implementing a cross-country survey from Israel and Czechia, combines data regarding revealed preferences about work habits before and during the pandemic and stated intentions data regarding anticipated work patterns when life returns to "normal" after the pandemic. Two models were used for the data analysis, one addressing factors that affect the increased/decreased teleworking trend and the other addressing factors that affect the frequency of actual commutes. The results reveal that most respondents (62% in Israel and 68% in Czechia) will maintain the same telecommuting/working from home balance. About 19% of respondents in both countries expressed their intention to reduce the number of commuting days, while 6% stated they would increase out-of-home days. However, these estimates rely only on workers' expectations not accounting for employers' point of view and other constraints they may have. Not accounting for potential bias, a moderate reduction of 6.5% and 8.7% (in Israel and in Czechia, respectively) in the number of commuting trips is expected in the post-pandemic era. The anticipated decrease in commuting days is accompanied by an increase in teleworking: from 10% to 14% among those who work more than 20 h a week (in both countries) and a drop in the rate of those who telework five hours or less a week (down from 73% to 63% in Israel and from 76% to 70% in Czechia). Self-employment, travel time to work, working solely on premise during the lockdown, and personal preferences regarding telework versus working away from home were found to significantly contribute to a decrease in the number of commuting days and to an increase in teleworking. An interesting finding is the high probability of increased teleworking among people who teleworked for the first time during the lockdown or who increased their teleworking time during the lockdown. This indicates that the teleworking experience due to the pandemic has enabled some people to view working from home as viable. Although, overall, the change in working habits does not seem dramatic, our results suggest that hybrid schemes for combining on premise and telework are expected to be adopted by some sectors.
ABSTRACT
Ordinal regression methods are widely used to predict the ordered labels of data, among which support vector ordinal regression (SVOR) methods are popular because of their good generalization. In many realistic circumstances, data are collected by a distributed network. In order to protect privacy or due to some practical constraints, data cannot be transmitted to a center for processing. However, as far as we know, existing SVOR methods are all centralized. In the above situations, centralized methods are inapplicable, and distributed methods are more suitable choices. In this paper, we propose a distributed SVOR (dSVOR) algorithm. First, we formulate a constrained optimization problem for SVOR in distributed circumstances. Since there are some difficulties in solving the problem with classical methods, we used the random approximation method and the hinge loss function to transform the problem into a convex optimization problem with constraints. Then, we propose subgradient-based algorithm dSVOR to solve it. To illustrate the effectiveness, we theoretically analyze the consensus and convergence of the proposed method, and conduct experiments on both synthetic data and a real-world example. The experimental results show that the proposed dSVOR could achieve close performance to that of the corresponding centralized method, which needs all the data to be collected together.
ABSTRACT
BACKGROUND: Protein protein interactions (PPIs) are essential to most of the biological processes. The prediction of PPIs is beneficial to the understanding of protein functions and thus is helpful to pathological analysis, disease diagnosis and drug design etc. As the amount of protein data is growing fast in the post genomic era, high-throughput experimental methods are expensive and time-consuming for the prediction of PPIs. Thus, computational methods have attracted researcher's attention in recent years. A large number of computational methods have been proposed based on different protein sequence encoders. RESULTS: Notably, the confidence score of a protein sequence pair could be regarded as a kind of measurement to PPIs. The higher the confidence score for one protein pair is, the more likely the protein pair interacts. Thus in this paper, a deep learning framework, called ordinal regression and recurrent convolutional neural network (OR-RCNN) method, is introduced to predict PPIs from the perspective of confidence score. It mainly contains two parts: the encoder part of protein sequence pair and the prediction part of PPIs by confidence score. In the first part, two recurrent convolutional neural networks (RCNNs) with shared parameters are applied to construct two protein sequence embedding vectors, which can automatically extract robust local features and sequential information from the protein pairs. Based on it, the two embedding vectors are encoded into one novel embedding vector by element-wise multiplication. By taking the ordinal information behind confidence score into consideration, ordinal regression is used to construct multiple sub-classifiers in the second part. The results of multiple sub-classifiers are aggregated to obtain the final confidence score. Following that, the existence of PPIs is determined by the confidence score. We set a threshold [Formula: see text], and say the interaction exists between the protein pair if its confidence score is bigger than [Formula: see text]. CONCLUSIONS: We applied our method to predict PPIs on data sets S. cerevisiae and Homo sapiens. Through experimental verification, our method outperforms state-of-the-art PPI prediction models.
Subject(s)
Neural Networks, Computer , Saccharomyces cerevisiae , Amino Acid Sequence , Humans , Proteins/geneticsABSTRACT
Regularization techniques such as the lasso (Tibshirani 1996) and elastic net (Zou and Hastie 2005) can be used to improve regression model coefficient estimation and prediction accuracy, as well as to perform variable selection. Ordinal regression models are widely used in applications where the use of regularization could be beneficial; however, these models are not included in many popular software packages for regularized regression. We propose a coordinate descent algorithm to fit a broad class of ordinal regression models with an elastic net penalty. Furthermore, we demonstrate that each model in this class generalizes to a more flexible form, that can be used to model either ordered or unordered categorical response data. We call this the elementwise link multinomial-ordinal (ELMO) class, and it includes widely used models such as multinomial logistic regression (which also has an ordinal form) and ordinal logistic regression (which also has an unordered multinomial form). We introduce an elastic net penalty class that applies to either model form, and additionally, this penalty can be used to shrink a non-ordinal model toward its ordinal counterpart. Finally, we introduce the R package ordinalNet, which implements the algorithm for this model class.
ABSTRACT
Thermal-stress events that cause coral bleaching and mortality have recently increased in frequency and severity. Yet few studies have explored conditions that moderate coral bleaching. Given that high light and high ocean temperature together cause coral bleaching, we explore whether corals at turbid localities, with reduced light, are less likely to bleach during thermal-stress events than corals at other localities. We analyzed coral bleaching, temperature, and turbidity data from 3,694 sites worldwide with a Bayesian model and found that Kd 490, a measurement positively related to turbidity, between 0.080 and 0.127 reduced coral bleaching during thermal-stress events. Approximately 12% of the world's reefs exist within this "moderating turbidity" range, and 30% of reefs that have moderating turbidity are in the Coral Triangle. We suggest that these turbid nearshore environments may provide some refuge through climate change, but these reefs will need high conservation status to sustain them close to dense human populations.
Subject(s)
Anthozoa , Animals , Bayes Theorem , Climate Change , Coral Reefs , Hot Temperature , TemperatureABSTRACT
Continuous response variables are often transformed to meet modeling assumptions, but the choice of the transformation can be challenging. Two transformation models have recently been proposed: semiparametric cumulative probability models (CPMs) and parametric most likely transformation models (MLTs). Both approaches model the cumulative distribution function and require specifying a link function, which implicitly assumes that the responses follow a known distribution after some monotonic transformation. However, the two approaches estimate the transformation differently. With CPMs, an ordinal regression model is fit, which essentially treats each continuous response as a unique category and therefore nonparametrically estimates the transformation; CPMs are semiparametric linear transformation models. In contrast, with MLTs, the transformation is parameterized using flexible basis functions. Conditional expectations and quantiles are readily derived from both methods on the response variable's original scale. We compare the two methods with extensive simulations. We find that both methods generally have good performance with moderate and large sample sizes. MLTs slightly outperformed CPMs in small sample sizes under correct models. CPMs tended to be somewhat more robust to model misspecification and outcome rounding. Except in the simplest situations, both methods outperform basic transformation approaches commonly used in practice. We apply both methods to an HIV biomarker study.