Representative Dataset Research Articles

Leveraging Imagery Data with Spatial Point Prior for Weakly Semi-supervised 3D Object Detection

Training high-accuracy 3D detectors necessitates massive labeled 3D annotations with 7 degree-of-freedom, which is laborious and time-consuming. Therefore, the form of point annotations is proposed to offer significant prospects for practical applications in 3D detection, which is not only more accessible and less expensive but also provides strong spatial information for object localization. In this paper, we empirically discover that it is non-trivial to merely adapt Point-DETR to its 3D form, encountering two main bottlenecks: 1) it fails to encode strong 3D prior into the model, and 2) it generates low-quality pseudo labels in distant regions due to the extreme sparsity of LiDAR points. To overcome these challenges, we introduce Point-DETR3D, a teacher-student framework for weakly semi-supervised 3D detection, designed to fully capitalize on point-wise supervision within a constrained instance-wise annotation budget. Different from Point-DETR which encodes 3D positional information solely through a point encoder, we propose an explicit positional query initialization strategy to enhance the positional prior. Considering the low quality of pseudo labels at distant regions produced by the teacher model, we enhance the detector's perception by incorporating dense imagery data through a novel Cross-Modal Deformable RoI Fusion (D-RoI). Moreover, an innovative point-guided self-supervised learning technique is proposed to allow for fully exploiting point priors, even in student models. Extensive experiments on representative nuScenes dataset demonstrate our Point-DETR3D obtains significant improvements compared to previous works. Notably, with only 5% of labeled data, Point-DETR3D achieves over 90% performance of its fully supervised counterpart.

Genomic surveillance of SARS-CoV-2 in North Africa: 4 years of GISAID data sharing

ObjectivesThis study aimed to construct geographically, temporally, and epidemiologically representative data sets for SARS-CoV-2 in North Africa, focusing on Variants of Concern (VOCs), Variants of Interest (VOIs), and Variants Under Monitoring (VUMs). MethodsSARS-CoV-2 genomic sequences and metadata from the EpiCoV database via the Global Initiative on Sharing All Influenza Data platform were analyzed. Data analysis included cases, deaths, demographics, patient status, sequencing technologies, and variant analysis. ResultsA comprehensive analysis of 10,783 viral genomic sequences from six North African countries revealed notable insights. SARS-CoV-2 sampling methods lack standardization, with a majority of countries lacking clear strategies. Over 59% of analyzed genomes lack essential clinical and demographic metadata, including patient age, sex, underlying health conditions, and clinical outcomes, which are essential for comprehensive genomic analysis and epidemiological studies, as submitted to the Global Initiative on Sharing All Influenza Data. Morocco reported the highest number of confirmed COVID-19 cases (1,272,490), whereas Tunisia leads in reported deaths (29,341), emphasizing regional variations in the pandemic's impact. The GRA clade emerged as predominant in North African countries. The lineage analysis showcased a diversity of 190 lineages in Egypt, 26 in Libya, 121 in Tunisia, 90 in Algeria, 146 in Morocco, and 10 in Mauritania. The temporal dynamics of SARS-CoV-2 variants revealed distinct waves driven by different variants. ConclusionsThis study contributes valuable insights into the genomic landscape of SARS-CoV-2 in North Africa, highlighting the importance of genomic surveillance in understanding viral dynamics and informing public health strategies.

Monitoring solutions for remote locations: A data gathering approach for remote nature-based solution sites

Nature-based solutions have gained popularity as an approach to tackling hydro-meteorological hazards (HMHs) in both urban and rural settings. Despite this popularity, challenges persist regarding the evidence base for their effectiveness and data scarcity at the feature or site scale. Flood modelling is a common approach to quantifying the effectiveness of NbS; however, the accuracy of these models heavily depends on the accuracy of the DEM, land cover, and hydraulic/hydrological data utilised. Remote and rural settings often face data scarcity due to the challenging nature of data collection, and insufficient funding for monitoring. Additionally, NbS features vary in size and scale, with many being small (<1 m in width), posing challenges for accurate representation in national LiDAR datasets. Technological advancements in remote sensing technologies, such as unmanned aerial vehicles, handheld LiDAR, and GPS-GNSS, offer opportunities to gather high-resolution, high-accuracy data in these challenging locations. This article proposes a methodological framework for collecting elevation data at remote NbS sites that can tackle areas affected by both sparse and dense vegetation cover. This approach proves valuable in both pre-NbS implementation, through facilitating NbS opportunity and environmental risk identification, and post-NbS implementation, through aiding in geo-spatial feature location, improving existing DEM data for flood modelling, and monitoring temporal changes.

Catching Fire: How National Humiliation Spreads Hostile Foreign Policy Preferences on Chinese Social Media

Abstract Research shows that emotions alter preferences, which are central to many models of political choice. However, how can theories that explain individual-level preference change explain policy outcomes, which usually take place at higher levels of analysis? I outline three competing approaches to this question, ultimately building on the third approach, which argues that emotions can spread through emotional contagion within identity groups. This implies emotions can broadly shift preferences towards or away from conflict during crises. It also explains how identities, which are relatively constant over the medium term, can suddenly become fault lines in a conflict. I find supporting evidence using a large (more than 1.6 billion posts) nationally representative data set of Chinese social media posts. Users’ national humiliation posts lead their followers to become more likely to make posts about national humiliation, suggesting contagion, and about using military force and maintaining territorial disputes, suggesting preference change.

Is social capital durable?: How family social bonds influence college enrollment and completion.

A large literature demonstrates that social capital has positive effects on outcomes for children, but we know little about whether social capital is durable, i.e., whether its effects persist long after its creation. We use two nationally representative data sets of U.S. high school students and structural equation modeling designed for binomial outcomes to examine the durability of returns to social capital created in the family on both college enrollment and college completion. Controlling for selected school characteristics, race, family, SES and other factors, results suggest that family social capital continues to have strong associations with outcomes increasingly distant from its creation. Family SES has a smaller but positive effect on both college enrollment and college completion. These findings suggest that social capital can be a durable good if formed in the family, and that family SES is also influential.

Frailty phenotype as mediator between systemic inflammation and osteoporosis and fracture risks: A prospective study.

Systemic inflammation and frailty have been implicated in osteoporosis (OP) and fracture risks; however, existing evidence remains limited and inconclusive. This study aimed to assess the associations of systemic inflammation and frailty phenotype with incident OP and fracture and to evaluate the mediating role of frailty phenotype. The present study analysed data from the UK Biobank, a comprehensive and representative dataset encompassing over 500000 individuals from the general population. Baseline peripheral blood cell counts were employed to calculate the systemic inflammation markers, including neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR) and systemic immune-inflammation index (SII). Frailty phenotype was assessed using five criteria, defined as frail (≥3 items met), pre-frail (1-2 items met) and non-frail (0 items met). OP and fracture events were confirmed through participants' health-related records. Multivariable linear and Cox regression models were utilized, along with mediation analysis. Increased systemic inflammation was associated with increased risks of OP and fracture. The corresponding hazard ratios and 95% confidence intervals (CIs) for OP risk per standard deviation increase in the log-transformed NLR, PLR and SII were 1.113 (1.093-1.132), 1.098 (1.079-1.118) and 1.092 (1.073-1.111), and for fracture risk, they were 1.066 (1.051-1.082), 1.059 (1.044-1.075) and 1.073 (1.058-1.089), respectively. Compared with the non-frail individuals, the pre-frail and frail ones showed an elevated OP risk by 21.2% (95% CI: 16.5-26.2%) and 111.0% (95% CI: 98.1-124.8%), respectively, and an elevated fracture risk by 6.1% (95% CI: 2.8-9.5%) and 38.2% (95% CI: 30.7-46.2%), respectively. The systemic inflammation level demonstrated a positive association with frailty, with β (95% CI) of 0.034 (0.031-0.037), 0.026 (0.023-0.029) and 0.008 (0.005-0.011) in response to per standard deviation increment in log-transformed SII, NLR and PLR, respectively. The frailty phenotype mediated the association between systemic inflammation and OP/fracture risk. Subgroup and sensitivity analyses confirmed the robustness of these findings. Systemic inflammation and frailty phenotype are independently linked to increased risks of OP and fracture. The frailty phenotype partially mediates the association between systemic inflammation and osteoporotic traits. These results highlight the significance of interventions targeting systemic inflammation and frailty in OP and fracture prevention and management.

Change in employment status and its causal effect on suicidal ideation and depressive symptoms: A marginal structural model with machine learning algorithms.

This study aimed to assess the causal effect of a change in employment status on suicidal ideation and depressive symptoms by applying marginal structural models (MSM) with machine-learning (ML) algorithms. We analyzed data from the 8-15th waves (2013-2020) of the Korean Welfare Panel Study, a nationally representative longitudinal dataset. Our analysis included 13 294 observations from 3621 participants who had standard employment at baseline (2013-2019). Based on employment status at follow-up year (2014-2020), respondents were classified into two groups: (i) maintained standard employment (reference group), (ii) changed to non-standard employment. Suicidal ideation during the past year and depressive symptoms during the past week were assessed through self-report questionnaire. To apply the ML algorithms to the MSM, we conducted eight ML algorithms to build the propensity score indicating a change in employment status. Then, we applied the MSM to examine the causal effect by using inverse probability weights calculated based on the propensity score from ML algorithms. The random forest algorithm performed best among all algorithms, showing the highest area under the curve 0.702, 95% confidence interval (CI) 0.686-0.718. In the MSM with the random forest algorithm, workers who changed from standard to non-standard employment were 2.07 times more likely to report suicidal ideation compared to those who maintained standard employment (95% CI 1.16-3.70). A similar trend was observed in the analysis of depressive symptoms. This study found that a change in employment status could lead to a higher risk of suicidal ideation and depressive symptoms.

A Spatiotemporal Enhanced SMAP Freeze/Thaw Product (1980–2020) over China and Its Preliminary Analyses

The soil freeze/thaw (FT) state has emerged as a critical role in the ecosystem, hydrological, and biogeochemical processes, but obtaining representative soil FT state datasets with a long time sequence, fine spatial resolution, and high accuracy remains challenging. Therefore, we propose a decision-level spatiotemporal data fusion algorithm based on Convolutional Long Short-Term Memory networks (ConvLSTM) to expand the SMAP-enhanced L3 landscape freeze/thaw product (SMAP_E_FT) temporally. In the algorithm, the Freeze/Thaw Earth System Data Record product (ESDR_FT) is sucked in the ConvLSTM and fused with SMAP_E_FT at the decision level. Eight predictor datasets, i.e., soil temperature, snow depth, soil moisture, precipitation, terrain complexity index, area of open water data, latitude and longitude, are used to train the ConvLSTM. Direct validation using six dense observation networks located in the Genhe, Maqu, Naqu, Pali, Saihanba, and Shandian river shows that the fusion product (ConvLSTM_FT) effectively absorbs the high accuracy characteristics of ESDR_FT and expands SMAP_E_FT with an overall average improvement of 2.44% relative to SMAP_E_FT, especially in frozen seasons (averagely improved by 7.03%). The result from indirect validation based on categorical triple collocation also shows that ConvLSTM_FT performs stable regardless of land cover types, climate types, and terrain complexity. The findings, drawn from preliminary analyses on ConvLSTM_FT from 1980 to 2020 over China, suggest that with global warming, most parts of China suffer from different degrees of shortening of the frozen period. Moreover, in the Qinghai–Tibet region, the higher the permafrost thermal stability, the faster the degradation rate.

Incomplete Multiview Clustering via Semidiscrete Optimal Transport for Multimedia Data Mining in IoT

With the wide deployment of the Internet of Things (IoT), large volumes of incomplete multiview data that violates data integrity is generated by various applications, which inevitably produces negative impacts on the quality of service of IoT systems. Incomplete multiview clustering (IMC), as an essential technique of data processing, has the potential for mining patterns of incomplete IoT data. However, previous methods utilize notion-strong distances that can only measure differences between distributions at the overlap of data manifolds in fusing complementary information of data for pattern mining. They may suffer from biased estimation and information loss in capturing intrinsic structures of incomplete multiview data. To address these challenges, a semidiscrete multiview optimal transport (SD-MOT) is defined for IMC, which utilizes distances with weak notions to capture intrinsic structures of incomplete multiview data. Specifically, IMC is recast as an equivalent optimal transport between continuous incomplete multiview data and discrete clustering centroids, to avoid the strict assumption on overlap between manifolds in pattern mining. Then, SD-MOT is instantiated as a deep incomplete contrastive clustering network to remedy biased estimation and information loss on intrinsic structures of incomplete multiview data. Afterwards, a variational solution to SD-MOT is derived to effectively train the network parameters for pattern mining. Finally, extensive experiments on four representative incomplete multiview datasets verify the superiority of SD-MOT in comparison with nine baseline methods.

Spatial-Pooling-Based Graph Attention U-Net for Hyperspectral Image Classification

In recent years, graph convolutional networks (GCNs) have attracted increasing attention in hyperspectral image (HSI) classification owing to their exceptional representation capabilities. However, the high computational requirements of GCNs have led most existing GCN-based HSI classification methods to utilize superpixels as graph nodes, thereby limiting the spatial topology scale and neglecting pixel-level spectral–spatial features. To address these limitations, we propose a novel HSI classification network based on graph convolution called the spatial-pooling-based graph attention U-net (SPGAU). Specifically, unlike existing GCN models that rely on fixed graphs, our model involves a spatial pooling method that emulates the region-growing process of superpixels and constructs multi-level graphs by progressively merging adjacent graph nodes. Inspired by the CNN classification framework U-net, SPGAU’s model has a U-shaped structure, realizing multi-scale feature extraction from coarse to fine and gradually fusing features from different graph levels. Additionally, the proposed graph attention convolution method adaptively aggregates adjacency information, thereby further enhancing feature extraction efficiency. Moreover, a 1D-CNN is established to extract pixel-level features, striking an optimal balance between enhancing the feature quality and reducing the computational burden. Experimental results on three representative benchmark datasets demonstrate that the proposed SPGAU outperforms other mainstream models both qualitatively and quantitatively.

Association between self-reported pain severity and characteristics of United States adults (age ≥50 years) who used opioids.

The aim of this study was to assess the associations between the characteristics of United States (US) adults (≥50 years) who used opioids and self-reported pain severity using a nationally representative dataset. This retrospective cross-sectional database study used 2019 Medical Expenditure Panel Survey data to identify US adults aged ≥50 years with self-reported pain within the past 4 weeks and ≥1 opioid prescription within the calendar year (n = 1,077). Weighted multivariable logistic regression analysis modeled associations between various characteristics and self-reported pain severity (quite a bit/extreme vs less/moderate pain). The adjusted logistic regression model indicated that greater odds of reporting quite a bit/extreme pain was associated with the following: age 50-64 vs ≥65 (adjusted odds ratio [AOR] = 1.76; 95% confidence interval [CI] = 1.22-2.54), non-Hispanic vs Hispanic (AOR = 2.0; CI = 1.18-3.39), unemployed vs employed (AOR = 2.01; CI = 1.33-3.05), no health insurance vs private insurance (AOR = 6.80; CI = 1.43-32.26), fair/poor vs excellent/very good/good health (AOR = 3.10; CI = 2.19-4.39), fair/poor vs excellent/very good/good mental health (AOR = 2.16; CI = 1.39-3.38), non-smoker vs smoker (AOR = 1.80; CI = 1.19-2.71), and instrumental activity of daily living, yes vs no (AOR = 2.27; CI = 1.30-3.96). Understanding the several characteristics associated with pain severity in US adults ≥50 years who used an opioid may help transform healthcare approaches to prevention, education, and management of pain severity in later life.

A novel approach for intelligent diagnosis and grading of diabetic retinopathy

Diabetic retinopathy (DR) is a severe ocular complication of diabetes that can lead to vision damage and even blindness. Currently, traditional deep convolutional neural networks (CNNs) used for DR grading tasks face two primary challenges: (1) insensitivity to minority classes due to imbalanced data distribution, and (2) neglecting the relationship between the left and right eyes by utilizing the fundus image of only one eye for training without differentiating between them. To tackle these challenges, we proposed the DRGCNN (DR Grading CNN) model. To solve the problem caused by imbalanced data distribution, our model adopts a more balanced strategy by allocating an equal number of channels to feature maps representing various DR categories. Furthermore, we introduce a CAM-EfficientNetV2-M encoder dedicated to encoding input retinal fundus images for feature vector generation. The number of parameters of our encoder is 52.88 M, which is less than RegNet_y_16gf (80.57 M) and EfficientNetB7 (63.79 M), but the corresponding kappa value is higher. Additionally, in order to take advantage of the binocular relationship, we input fundus retinal images from both eyes of the patient into the network for features fusion during the training phase. We achieved a kappa value of 86.62% on the EyePACS dataset and 86.16% on the Messidor-2 dataset. Experimental results on these representative datasets for diabetic retinopathy (DR) demonstrate the exceptional performance of our DRGCNN model, establishing it as a highly competitive intelligent classification model in the field of DR. The code is available for use at https://github.com/Fat-Hai/DRGCNN.

Determinants of organic soil fertilization methods use over time and in the face of climate vulnerability

Implementation of environment-friendly soil organic matter (SOM) enhancing technologies (SOMET) is crucial for addressing soil degradation. This study aims to examine the usage status of SOMET (i.e., organic fertilizer, manure, and compost) from the dimension of long-term non-usage, dis-usage, late-usage and long-term usage and identify the drivers of these four behaviors of usage. We utilized national representative datasets of 1659 Bangladeshi rice-farmers for the periods of 2013, 2016 and 2020 with climate hazards data. Households were categorized into long-term non-user (48%) who does not practice SOMET in any of three survey years, dis-user (29%) as the households who abandon SOMET after practicing a period of time, late-user (20%) for the households who apply SOMET a few years later than their peer, long-term user (3%) as the households who practice SOMET for three survey years. Ordered logit model was used to quantify drivers of likelihood of being the four above defined categories. SOMET use has been found to be highly constrained by climate hazards. Flood depth, salinity, heavy rainfall, storm and cyclone vulnerability decrease likelihood of long-term use and late-use while increase the probability of long-term non-use and dis-use (p≤0.01). Alternatively, the likelihood of long-term non-use and dis-use are 12% and 4% lower in drought-prone areas. Increasing drought experiences increase the probability of long-term use and late-use (p≤0.01). Higher SOM level decreases long-term non-use but induces dis-use. Other major drivers of SOMET long-term use are older household head, more educated women in households, larger farm-size, and higher livestock values. Considering these factors in developing and implementing policies could be instrumental in promoting SOMET application at farm-level. Long-term use is context-specific with various climate and socio-economic factors, thus, designing policies and strategies should emphasize contextual variations to promote usage continuities.

Local Crime and Early Marriage: Evidence from India

This paper analyses whether living in a locality with high crime against women affects the probability of early marriage—that is, marriage before the legal age of marriage of girls. Using a nationally representative longitudinal data set and tackling the potential endogeneity of local crime rates, we find that perceived crime against women in the locality significantly increases the likelihood of early marriage of girls, while there is no such effect on boys of comparable age group. We also find no such effect of gender-neutral crimes (such as theft and robbery) on the likelihood of early marriage of girls. Moreover, we find that the relationship holds only in conservative households where the purdah system is practised, and also in the northern region of India, where patriarchal culture and gender norms are stronger than in the southern region. A sensitivity analysis assessing the potential impact of unobservable confounders suggests that our estimates are unlikely to be affected by omitted variable bias.

Back to normal? ESL Instructors' Post-Pandemic Perceptions of Virtual Learning

Back to normal? ESL Instructors' Post-Pandemic Perceptions of Virtual Learning

Synthetic Aperture Radar Image Compression Based on Low-Frequency Rejection and Quality Map Guidance

Synthetic Aperture Radar (SAR) images are widely utilized in the field of remote sensing. However, there is a limited body of literature specifically addressing the compression of SAR learning images. To address the escalating volume of SAR image data for storage and transmission, which necessitates more effective compression algorithms, this paper proposes a novel framework for compressing SAR images. Experimental validation is performed using a representative low-resolution Sentinel-1 dataset and the high-resolution QiLu-1 dataset. Initially, we introduce a novel two-stage transformation-based approach aimed at suppressing the low-frequency components of the input data, thereby achieving a high information entropy and minimizing quantization losses. Subsequently, a quality map guidance image compression algorithm is introduced, involving the fusion of the input SAR images with a target-aware map. This fusion involves convolutional transformations to generate a compact latent representation, effectively exploring redundancies between focused and non-focused areas. To assess the algorithm’s performance, experiments are carried out on both the low-resolution Sentinel-1 dataset and the high-resolution QiLu-1 dataset. The results indicate that the low-frequency suppression algorithm significantly outperforms traditional processing algorithms by 3–8 dB when quantifying the input data, effectively preserving image features and improving image performance metrics. Furthermore, the quality map guidance image compression algorithm demonstrates a superior performance compared to the baseline model.

Where and How to Transfer: Knowledge Aggregation-Induced Transferability Perception for Unsupervised Domain Adaptation.

Unsupervised domain adaptation without accessing expensive annotation processes of target data has achieved remarkable successes in semantic segmentation. However, most existing state-of-the-art methods cannot explore whether semantic representations across domains are transferable or not, which may result in the negative transfer brought by irrelevant knowledge. To tackle this challenge, in this paper, we develop a novel Knowledge Aggregation-induced Transferability Perception (KATP) for unsupervised domain adaptation, which is a pioneering attempt to distinguish transferable or untransferable knowledge across domains. Specifically, the KATP module is designed to quantify which semantic knowledge across domains is transferable, by incorporating transferability information propagation from global category-wise prototypes. Based on KATP, we design a novel KATP Adaptation Network (KATPAN) to determine where and how to transfer. The KATPAN contains a transferable appearance translation module T_A() and a transferable representation augmentation module T_R(), where both modules construct a virtuous circle of performance promotion. T_A() develops a transferability-aware information bottleneck to highlight where to adapt transferable visual characterizations and modality information; T_R() explores how to augment transferable representations while abandoning untransferable information, and promotes the translation performance of T_A() in return. Experiments on several representative datasets and a medical dataset support the state-of-the-art performance of our model.

Optimising brain age estimation through transfer learning: A suite of pre-trained foundation models for improved performance and generalisability in a clinical setting.

Estimated age from brain MRI data has emerged as a promising biomarker of neurological health. However, the absence of large, diverse, and clinically representative training datasets, along with the complexity of managing heterogeneous MRI data, presents significant barriers to the development of accurate and generalisable models appropriate for clinical use. Here, we present a deep learning framework trained on routine clinical data (N up to 18,890, age range 18-96 years). We trained five separate models for accurate brain age prediction (all with mean absolute error ≤4.0 years, R2 ≥ .86) across five different MRI sequences (T2 -weighted, T2 -FLAIR, T1 -weighted, diffusion-weighted, and gradient-recalled echo T2 *-weighted). Our trained models offer dual functionality. First, they have the potential to be directly employed on clinical data. Second, they can be used as foundation models for further refinement to accommodate a range of other MRI sequences (and therefore a range of clinical scenarios which employ such sequences). This adaptation process, enabled by transfer learning, proved effective in our study across a range of MRI sequences and scan orientations, including those which differed considerably from the original training datasets. Crucially, our findings suggest that this approach remains viable even with limited data availability (as low as N = 25 for fine-tuning), thus broadening the application of brain age estimation to more diverse clinical contexts and patient populations. By making these models publicly available, we aim to provide the scientific community with a versatile toolkit, promoting further research in brain age prediction and related areas.

Application of the Sleep C.A.L.M. Tool for Assessing Nocturia in a Large Nationally Representative Cohort.

Nocturia significantly impacts patients' quality of life but remains insufficiently evaluated and treated. The "Sleep C.A.L.M." system categorizes the factors thought to collectively reflect most underlying causes of nocturia (Sleep disorders, Comorbidities, Actions [i.e., modifiable patient behaviors such as excess fluid intake], Lower urinary tract dysfunction, and Medications). The purpose of this study was to assess the association of nocturia with the Sleep C.A.L.M. categories using a nationally representative dataset. Retrospective analysis of the National Health and Nutrition Examination Survey from 2013/14-2017/18 cycles was conducted. Pertinent questionnaire, laboratory, dietary, and physical examination data were used to ascertain the presence of Sleep C.A.L.M. categories in adults ≥20 years of age. Nocturia was defined as ≥2 nighttime voids. A total of 12,274 included subjects were included (51.6% female; median age, 49.0 years [interquartile range, 34.0-62.0 years]; 27.6% nocturia). Among subjects with nocturia, the prevalence of 0, ≥1, and ≥2 Sleep C.A.L.M. categories was 3.5% (95% confidence interval [CI], 2.8%-4.4%), 96.5% (95% CI, 95.6%-97.2%), and 81.2% (95% CI, 78.9%-83.3%), respectively. Compared to those with 0-1 Sleep C.A.L.M. categories, the adjusted odds of nocturia in subjects with 2, 3, and 4-5 Sleep C. A.L.M. categories were 1.77 (95% CI, 1.43-2.21), 2.33 (1.89-2.87), and 3.49 (2.81-4.35), respectively (P<0.001). Similar trends were observed for most age and sex subgroups. When assessed individually, each of the 5 Sleep C.A.L.M. categories were independently associated with greater odds of nocturia, which likewise persisted across multiple age and sex subgroups. Sleep C.A.L.M. burden is associated with increased odds of nocturia in a dose-dependent fashion, and potentially a relevant means by which to organize the underlying etiologies for nocturia among community-dwelling adults.

A unifying perspective on non-stationary kernels for deeper Gaussian processes

The Gaussian process (GP) is a popular statistical technique for stochastic function approximation and uncertainty quantification from data. GPs have been adopted into the realm of machine learning (ML) in the last two decades because of their superior prediction abilities, especially in data-sparse scenarios, and their inherent ability to provide robust uncertainty estimates. Even so, their performance highly depends on intricate customizations of the core methodology, which often leads to dissatisfaction among practitioners when standard setups and off-the-shelf software tools are being deployed. Arguably, the most important building block of a GP is the kernel function, which assumes the role of a covariance operator. Stationary kernels of the Matérn class are used in the vast majority of applied studies; poor prediction performance and unrealistic uncertainty quantification are often the consequences. Non-stationary kernels show improved performance but are rarely used due to their more complicated functional form and the associated effort and expertise needed to define and tune them optimally. In this perspective, we want to help ML practitioners make sense of some of the most common forms of non-stationarity for Gaussian processes. We show a variety of kernels in action using representative datasets, carefully study their properties, and compare their performances. Based on our findings, we propose a new kernel that combines some of the identified advantages of existing kernels.