Search Results (675)

Pedestrian gender recognition plays an important role in pedestrian analysis and intelligent video applications, for example, in demographic statistics, soft biometric analysis, and context-aware person retrieval. However, it remains a challenging task owing to viewpoint variations, illumination changes, occlusions, and low image quality in real-world imagery. To address these issues, an effective adaptive decision fusion framework, termed the Decision Fusion Learning Network (DFLN), is proposed in this paper. The key novel aspect of DFLN is that it effectively explores both an appearance-centered view that emphasizes detailed texture and clothing information and a structure-centered view that captures rich contour and structural information for pedestrian gender recognition. To realize DFLN, a Parallel CNN Prediction Probability Learning Module (PCNNM) is first constructed to independently learn modality-specific probabilities from color image and edge maps. Subsequently, a learnable Decision Fusion Module (DFM) is designed to fuse the modality-specific probabilities and explore their complementary merits for realizing accurate pedestrian gender recognition. The DFM can be easily coupled with the PCNNM, forming an end-to-end decision fusion learning framework that simultaneously learns the feature representations and carries out adaptive decision fusion. Experiments on two pedestrian benchmark datasets, named PETA and PA-100K, show that DFLN achieves competitive or superior performance compared with several state-of-the-art pedestrian gender recognition methods. Extensive experimental analysis further confirms the effectiveness of the proposed decision fusion strategy and its favorable generalization ability under domain shift. Full article

Background: Incorporating sex as a biological variable (SBV) is recognized as essential for improving the reliability, reproducibility, and generalizability of pharmacological research. This principle is codified in international policies and guidelines, yet implementation remains uneven, especially in phytomedicine. Phytomedicines are a major component of healthcare worldwide, with 65% of the global population relying on them in both regulated and traditional contexts. Globally, phytomedicines are used by males, females, intersex and non-cis gender persons, all of whom may present specific safety and efficacy considerations and warrant full inclusion in pre-clinical to clinical research pipelines. However, in contemporary settings, phytomedicine lags in SBV best practices relative to Western allopathic standards for research design. Methods: We conducted a non-systematic review and in silico data mining to quantify sex/gender representation in recent preclinical and clinical phytomedicine studies, complemented by targeted case studies of sexually dimorphic safety/efficacy. We also summarize the historical role of women and gender-diverse people as users and providers within Traditional and Integrative Medical Systems (TIMSs). Results: Across rodent and human studies, females are under-represented relative to males, and sex is rarely reported for cell lines. Intentional inclusion of intersex and other gender-diverse populations is largely absent. Case studies illustrate plausible sex-associated differences in pharmacokinetics, pharmacodynamics, and adverse event profiles. TIMSs historically address women’s health needs and include substantial participation by female practitioners; however, contemporary SBV practices remain less standardized than in Western allopathic pipelines. Conclusions: SBV integration in phytomedicine is needed to strengthen safety, efficacy, and regulatory-grade evidence. Practical barriers include legacy datasets without sex metadata, limited intersex animal models, and uneven resources across settings. We outline feasible, stepwise practices to improve SBV adoption in a manner compatible with TIMS contexts and recommend expanding current guidelines to better support diverse research environments while maintaining scientific rigor. Full article

Student attrition remains a persistent challenge in higher education and is shaped by interacting socioeconomic, academic, institutional, and wellbeing-related mechanisms. Although learning analytics and educational data mining increasingly support early-warning and intervention workflows, dataset reuse is often limited by incomplete documentation and inconsistent variable definitions. This Data Descriptor presents a structured cross-sectional survey dataset on factors influencing student persistence at a Colombian public university campus (La Paz). Data were collected between August and December 2025 through an online questionnaire and subsequently cleaned to remove duplicate entries and personally identifiable information. The released dataset contains 333 student records and 33 variables covering demographics (e.g., age, gender, first-generation status), socioeconomic conditions (e.g., residential stratum, housing, financial aid), academic experience and satisfaction (multiple 1–5 Likert items), perceived dropout intention across personal/socioeconomic/academic domains, thematically coded open-ended items describing challenges and motives, and a self-allocation of 0–100 weights across three dropout-factor domains. We provide a machine-readable codebook, a transparent preprocessing description, and technical validation checks (value ranges, category consistency, and composite-score integrity). The dataset is intended to support reproducible retention research, equity-oriented analyses, and benchmarking of predictive models, while encouraging responsible reuse through privacy-preserving release practices and FAIR-aligned metadata, repository deposition, and versioning. Full article

Physical inactivity remains a major modifiable risk factor for non-communicable diseases and continues to exhibit marked socioeconomic and gender disparities in Latin America. Identifying robust and interpretable predictors of inactivity in nationally representative datasets is essential for informing public health strategies. This study compared a survey-weighted logistic regression model and an explainable machine learning approach (XGBoost) to predict physical inactivity among Chilean adults using data from the 2024 National Physical Activity and Sports Survey (ENAFyD; n = 5248). Models were evaluated on a stratified held-out test set (n = 1050) using weighted and unweighted area under the ROC curve (AUC), Brier scores, and calibration curves. Survey-weighted logistic regression achieved a weighted AUC of 0.801, while XGBoost achieved 0.797, demonstrating comparable discrimination. XGBoost showed marginally lower Brier scores, indicating slightly improved probabilistic calibration. Low socioeconomic status, female sex, lower monthly physical activity expenditure, limited facility access, and lower engagement with digital resources were consistently associated with higher inactivity risk. SHAP-style contribution analysis provided additional insight into feature-level influence within the machine learning framework. Overall, both approaches demonstrated similar predictive capacity, supporting the complementary use of classical regression and explainable machine learning for population-level physical inactivity research. Full article

Although sexuality education is delivered in schools, Chinese adolescents’ preferred sources may still be inconsistent with where they currently receive sex education. Based on two theories (Objectivism and Sex-Positivity) that emphasize the use of information and rational choice in seeking information with a desire to learn more, this explanatory study analyses survey data on sexual health topics, the sources Chinese adolescents used and preferred, and the gender differences in how they seek information on sexual topics. From 17 secondary schools, Chinese adolescents in Hong Kong, aged 14–18 (n = 4869), took a 51-question survey on sexual risks and sex education sources conducted by a local agency. Using the secondary dataset, a discrepancy score was computed by matching 15 actually used and 15 preferred sources for getting sexual knowledge listed in the survey. The discrepancy scores were shifted along the X-axis to eliminate negative values and create the dependent variable, ‘Discrepancy-S’, which ranged from 1 to 11, where 1 = no discrepancy, and 11 = wide discrepancy (Cronbach Alpha = 0.750). The higher the score, the higher the discrepancy. Regression results indicated that the youth’s prior coitus and different information sources (except school) could explain the “use–prefer” discrepancy. Although these adolescents regarded parents as the primary sex educators, most did not consult with their families. They preferred electronic media and peers as their top “go-to” choices. Sex education should come from sources that teenagers rely on and choose to access. Personal responsibility must be explicitly discussed in various sexual health sources as teens prepare for transitions to adulthood. Full article

Gender bias embedded in legal and professional texts perpetuates systemic inequality, yet research on bias identification and mitigation remains largely confined to English. Morphologically rich languages such as Greek, where grammatical gender pervades nouns, adjectives, pronouns, and participles, present unique challenges that existing approaches fail to address. This paper elaborates on a systematic methodology primarily focusing on identifying and mitigating gender bias in Greek-language job advertisements and legal documents. To accomplish that task, we define a taxonomy of nine gender bias rules tailored to the linguistic properties of Greek and construct domain-specific annotated datasets comprising 90 expert-curated few-shot examples across both textual domains. Using these resources, we employ XML-structured prompt engineering with in-context learning (ICL)and systematically compare three classes of models: (i) commercial large language models (LLMs), namely Claude Sonnet 4.5 and GPT-5.2, (ii) two open-weight small language models (SLMs), Mistral Small (24B) and Ministral (14B), and (iii) Llama Krikri (8B), a Greek-native language model built on Llama 3.1 and fine-tuned on high-quality Greek corpora. For each input text, the system identifies biased expressions, maps them to specific bias rules, provides explanations, and generates a fully corrected inclusive version. Our experiments reveal substantial performance disparities across model scales and linguistic specialization, with LLMs demonstrating superior contextual reasoning and SLMs exhibiting systematic over-correction and grammatical errors in Greek morphology. We further introduce a critical meta-rule addressing gender agreement with named entities to prevent spurious corrections in legal texts referencing identified individuals. The findings highlight the importance of model scale, language-specific adaptation, and carefully designed prompting strategies for bias mitigation in underrepresented languages. Full article

Given its substantial contribution to traffic accidents, one of the main goals of intelligent driver-assistance systems has become the detection and mitigation of driver fatigue to enhance driving safety and comfort. Among various approaches, vision-based facial analysis using deep learning has emerged as an effective and non-intrusive method for identifying driver drowsiness, as a key manifestation of fatigue. However, current drowsiness detection models do not account for demographic factors like gender, even though recent research has shown gender behavioral differences such as eye closure duration, blink frequency, yawning patterns, and facial muscle relaxation. In this paper, we present a fine-grained multi-stream transformer architecture that incorporates gender-awareness and shifted-windows attention for spatial feature fusion. Integrating gender embedding, by modulating the region-based features, allows the model to effectively learn gender-conditioned drowsiness features to minimize bias and diluted representations. Using the NTHU-DDD dataset, we evaluated two-stream and three-stream variants for gender-aware and gender-agnostic across three facial region contexts: the face region with a 20% margin, bare face region, and key facial regions (face, eyes, and mouth). A comprehensive ablation study was conducted to identify the most effective model setup. The results demonstrate that incorporating gender embedding improves detection performance, achieving an accuracy of 95.47% on the evaluation set. Moreover, using the proposed three-stream model (SWT-DD-3S) produced better results. Full article

This research examines the psychometric properties of the Italian Multidimensional Psychological Flexibility Inventory short form (MPFI-24), a measure of psychological flexibility/inflexibility. Study 1 investigated its factor structure, reliability and invariance (across gender, age, and mental health status) based on a dataset comprising 1542 participants (71% female, mean_age = 38.6 years, SD = 15.0). Study 2 reexamined the factorial structure in an independent sample (N = 728, 64.88% females, mean_age = 30.94 years, SD = 14.07), and assessed both convergent validity (with psychological flexibility/inflexibility measures) and concurrent validity (with distress and well-being measures). Confirmatory factor analyses demonstrated very good fit indices for a first-order model comprising the twelve psychological flexibility and inflexibility sub-processes. In addition, the model structured with two second-order factors—psychological flexibility and inflexibility—each defined by six core sub-processes, showed a good model fit. The Italian MPFI-24 also exhibited strong internal consistency and good convergent and concurrent validity. Measurement invariance was established for gender, age, and mental health status. The Italian MPFI-24 is a psychometrically sound instrument for evaluating psychological flexibility and inflexibility, along with their underlying sub-processes, in an Italian context. Full article

Based on data from 208 students involved in entrepreneurship studies at Tecnológico de Monterrey, Mexico, this paper examines whether prior experience with AI, expectations, gender, and age reinforce future AI use. To achieve this objective, we applied binary logistic regression with random oversampling to balance the dataset. We complemented it with additional model performance metrics, including the confusion matrix, sensitivity, specificity, and area under the ROC curve. The results show that prior experience with AI, age-related technology use, and positive expectations regarding AI are associated with a higher likelihood of reinforcing future AI use. In terms of gender, the results indicate a gender gap favoring women, who are more likely to use AI when they perceive greater utility and confidence, as well as a stronger desire to succeed. Full article

Moving beyond traditional sociodemographic models, this study investigates the psychometric drivers of lexical change. Using Swiss German as a case study, we compare historical data from the Sprachatlas der deutschen Schweiz (1939–1958) with a recent large-scale app-based survey (N = 1013) to quantify trajectories over the past century. We identify four distinct mechanisms: exogenous convergence (Schmetterling), endo-normative leveling (Rande), endogenous innovation and divergence (schlittschuhlaufen), and diachronic persistence (Stäge). For the locally rooted speakers in our dataset, structural analysis indicates that traditional variables carry less weight than expected. While age remains the primary vertical predictor, psychological factors outperform traditional variables (e.g., gender, social networks) in this environment of ubiquitous exposure. Multivariate models demonstrate that lexical choices are strongly influenced by individual disposition: traits such as agreeableness accelerate the adoption of supraregional forms, whereas a strong local identity functions as a “brake” against standardization. Ultimately, while macro-factors create the pressure for change, individual micro-factors determine whether it takes hold. A speaker’s attitude acts as a “filter” and their personality as a “gate,” deciding whether they accept or resist new forms. These findings challenge purely structural accounts, suggesting that for these locally rooter speakers, even without high physical mobility, lexical change is shaped by a psychometric architecture. Full article

Background: Warfarin pharmacogenomics is critical due to its narrow therapeutic index and significant interpatient variability. While machine learning (ML) can predict anticoagulation control status (ACS), its “black-box” nature limits clinical translatability. Explainable Artificial Intelligence (XAI) addresses this by providing interpretable insights. This study applied ML and XAI to a warfarin pharmacogenomic dataset to predict poor ACS and explain model decisions. Methods: A post hoc analysis was conducted on a cross-sectional dataset of 232 patients receiving warfarin for ≥6 months. Data included age, gender, interacting drugs, SAMe-TT2R2 score, and genotypes for CYP2C9, VKORC1, and CYP4F2. Poor ACS was defined as time in therapeutic range (TTR) < 70%. The dataset was split into training (70%) and testing (30%) cohorts. Three models, Random Forest, XGBoost, and Logistic Regression, were developed and evaluated using AUC-ROC, sensitivity, and specificity. XAI techniques, including permutation importance and SHapley Additive exPlanations (SHAP), were employed for global and local interpretability. Results: Of 232 patients, 141 (60.8%) had poor ACS. XGBoost and Random Forest demonstrated comparable predictive accuracy (AUC-ROC: 0.67), outperforming Logistic Regression. Sensitivity was 0.83 and 0.79 for XGBoost and Random Forest, respectively. However, specificity was modest for both ensemble methods (Random Forest: 0.48; XGBoost: 0.41) and extremely low for Logistic Regression (0.04), indicating poor discrimination, particularly for identifying patients with adequate anticoagulation control. Globally, important predictors included the age, SAMe-TT2R2 score, CYP2C9 (*2/*2), female gender, and VKORC1 (C/T). XAI revealed predictions were primarily driven by VKORC1, CYP4F2, SAMe-TT2R2 scores, and drug interactions. Concordance between XAI predictions and actual ACS was 78% for adequate and 88.6% for poor ACS. SHAP analysis showed VKORC1 provided a stable risk signal (mean absolute SHAP: 1.44 ± 0.49 in concordant cases), while CYP2C9 was a high-variance, high-impact driver of discordance (mean SHAP: 3.44 ± 3.79 in discordant cases). Conclusions: ML models, particularly ensemble methods, show modest ability to predict poor warfarin control with limited ability to correctly identify patients with adequate control from our dataset. XAI transforms these models into interpretable tools, with SHAP analysis attributing predictions to specific genetic and clinical features. While predictive accuracy remains modest, this approach enhances transparency and provides a foundation for generating hypotheses that may ultimately support clinical decision-making in pharmacogenomic-guided warfarin therapy. Full article

Medical image segmentation, particularly for pathological diagnosis, faces challenges in leveraging patient clinical metadata that could enhance diagnostic accuracy. This study presents MMY-Net (Multimodal Y-shaped Network), a novel deep learning framework that effectively fuses patient metadata with pathological images for improved tumor segmentation performance. The proposed architecture incorporates a Text Processing Block (TPB) utilizing BERT for metadata feature extraction and a Text Encoding Block (TEB) for multi-scale fusion of textual and visual information. The network employs an Interlaced Sparse Self-Attention (ISSA) mechanism to capture both local and global dependencies while maintaining computational efficiency. Experiments were conducted on two open/public eyelid tumor datasets (Dataset 1: 112 WSIs for training/validation; Dataset 2: 107 WSIs as an independent test set) and the public Dataset 3 gland segmentation benchmark. For Dataset 1, 7989 H&E-stained patches (1024 × 1024, resized to 224 × 224) were extracted and split 7:2:1 (train:val:test); Dataset 2 was used exclusively for external validation. All images underwent Vahadane stain normalization. Training employed SGD (lr = 0.001), 1000 epochs, and a hybrid loss (cross-entropy + MS-SSIM + Lovász). Results show that integrating metadata—such as age and gender—significantly improves segmentation accuracy, even when metadata does not directly describe tumor characteristics. Ablation studies confirm the superiority of the proposed text feature extraction and fusion strategy. MMY-Net achieves state-of-the-art performance across all datasets, establishing a generalizable framework for multimodal medical image analysis. Full article

Asymmetries in data distributions and performance across subgroups can induce systematic unfairness in real-world systems. A variety of previous studies have significantly ameliorated the fairness of deep learning models; however, most of them necessarily require additional labels for sensitive attributes, (i.e., ethnicity and gender). Since sensitive attributes often correspond to personal information, collecting such labels can be restricted and may raise privacy concerns. Although recent work has sought to address these issues by training a model without sensitive attribute labels, we point out that it has limitations, as it assumes specific characteristics of sensitive attributes and is validated in simplistic, constrained environments. Therefore, we propose an Unsupervised Fairness-aware Framework (UFF) that trains a fair classification model without pre-defining the characteristics of the sensitive attributes. It includes branches that capture various types of biases and eliminates them through adversarial training. In various scenarios on benchmark datasets, (i.e., CelebA and UTK Face) for facial attribute classification, the proposed method significantly enhances fairness without assuming specific characteristics of sensitive attributes. Moreover, we introduce g-FAT, which is a new metric to measure generalized trade-off performances between classification accuracy and fairness. For example, on CelebA, ours reduces EO from 11.8 to 7.6 for malignant bias and from 15.6 to 9.6 for benign bias, while improving g-FAT from 80.7 to 84.9 and from 79.0 to 85.2, respectively. In terms of g-FAT, our method achieves the highest trade-off performance among the compared methods on the benchmarks. Full article

Banks often assume that higher credit limits increase customer default risk because greater exposure appears to imply greater vulnerability. This reasoning, however, conflates correlation with causation. Whether increasing a customer’s credit limit truly raises the likelihood of default remains an open empirical question that this work seeks to answer. We applied Bayesian causal inference to estimate the causal effect of credit limits on default probability. The analysis incorporated Directed Acyclic Graphs (DAGs) for causal structure, d-separation for identification, and Bayesian logistic regression using a dataset of 30,000 credit card holders in Taiwan (April–September 2005). Twenty-two confounding variables were adjusted for, covering demographics, repayment history, and billing and payment behavior. Continuous covariates were standardized, and posterior inference was performed using NUTS sampling with posterior predictive simulations to compute Average Treatment Effects (ATEs). We found that a one-standard-deviation increase in credit limit reduces default probability by 1.44 percentage points (94% HDI: [−2.0%, −1.0%]), corresponding to a 6.3% relative decline from the baseline default rate of 22.1%. The effect was consistent across demographic subgroups, with homogeneous treatment effects observed for age, education, and gender categories, and remained robust under sensitivity analysis addressing potential unmeasured confounding. The findings suggest that increasing credit limits can causally reduce default risk, likely by enhancing financial flexibility and lowering utilization ratios. These results have practical implications for credit policy design and motivate further investigation into mechanisms and applicability across broader lending environments. These estimates are explicitly interpreted as context-specific causal effects for a pre-crisis consumer credit environment, with external validity assessed conceptually rather than assumed. Full article

Fairness-aware classification with respect to sensitive attributes, such as gender and race, is one of the most important topics in machine learning. Although numerous studies have made outstanding progress through various approaches, one key limitation is that they necessarily require additional labels of sensitive attributes for training. This poses a significant challenge since sensitive attributes typically correspond to personal information. To this end, we propose a novel reweighting method that dynamically gives more weights to underrepresented groups across potential sensitive attributes. Without auxiliary networks or strong assumptions about sensitive attributes, the proposed method significantly improves fairness under various scenarios on benchmark datasets, outperforming the existing state-of-the-art methods. Full article