Search Results (3,119)

We propose a transformer-based neural network for predicting credit card default using raw multivariate credit data represented as a 2D time series, eliminating the need for manual feature engineering. Unlike existing state-of-the-art (SOTA) tree-based models that rely heavily on handcrafted features, our model leverages self-attention to extract latent temporal patterns directly from the raw data. Evaluated on two real-world datasets, our approach outperforms the popular LightGBM baselines and achieves performance on par with the leading ensemble methods. To further explore if our proposed model can enhance common ensemble methods, we incorporate it into an ensemble together with LightGBM. Experimental results show that the ensemble integrating our proposed transformer-based model outperforms existing ensemble approaches. Designed with deployment in mind, the model architecture is lightweight, generalizable, and maintainable, making it suitable for integration into real-world credit risk pipelines. Our results demonstrate strong practical relevance and a clear path towards scalable deployment in financial applications. In addition, we have built in an optional feature augmentation extension to the proposed model to facilitate hybrid adoption of our model by existing users who are accustomed to engineered features from domain expertise and industry practice. Hence, our model is user-friendly and can leverage hybrid learning to support both user-crafted and model-learned features to improve model performance and deployment. Full article

Proactive robotic assistance in human–robot collaboration (HRC) requires systems that can perceive evolving task contexts, anticipate user needs, and intervene appropriately without disrupting human workflow. We present the Agentic Unified Robotic Assistance (AURA) Framework, which couples Large Language Model (LLM) reasoning grounded by Standard Operating Procedures (SOPs) with a modular layer of specialized Intent, Motion, Perception, Sound, Affordance, and Performance Monitors that supply structured context to a central decision-making module, making the framework reconfigurable and auditable without retraining or re-prompting. We introduce a human-in-the-loop teleoperation data collection methodology and an offline evaluation scheme with an Appropriateness Score (A-Score) tailored to proactive intervention timing, and release a benchmark dataset of annotated multimodal HRC episodes containing workspace and robot wrist camera videos, robot joint states, and labeled intervention events. Across three tasks of varying complexity, we observe progressive gains in intent prediction and decision-making as the modules are supplied with richer grounded context (prior-state memory and tracked object locations), with Combined F1 rising by over 20 points between context-poor and context-rich conditions. The structured grounding allows lightweight multimodal backbones such as Gemini 3.1 Flash Lite to perform on par with heavier reasoning-tier models at roughly one-fifth the inference latency. Together, these contributions establish a scalable framework, benchmark, and evaluation methodology for advancing proactive robotic assistance in collaborative environments. Full article

As recent Multi-Layer Perceptron (MLP) mixer models have achieved state-of-the-art performance in time series forecasting, modeling each MLP-mixer as a separate expert within a mixture is expected to extend the representational capacity of the model, allowing each expert to be activated in response to time-varying inputs. However, extending MLP-mixers into a Mixture-of-Experts (MoE) architecture introduces a significant increase in the number of trainable parameters, rendering the model more challenging to train. To mitigate this problem, we propose a method that composes a fully trainable global expert and multiple non-trainable local experts. Specifically, our approach clones the weights of the global expert into the local experts and then modifies their weight distributions using moment learning, a recently proposed unconventional method for training neural networks. Concretely, each local expert is produced by applying moment-based transformations to a shared copy of the global expert’s weights, so that expert specialization is obtained without independently training the additional experts. Experimental results using a lightweight Time Series Mixer (TSMixer) architecture demonstrate that our method achieves performance competitive with fully trainable MoE counterparts, without introducing a significant increase in trainable parameters. Across multiple benchmark settings, the proposed model attains forecasting accuracy on par with, and in several cases favorable to, a fully trainable multi-expert baseline while adding only a small fraction of the extra trainable parameters that such a baseline requires, and this efficiency is further corroborated by measurements of memory footprint as well as an effect-size-based assessment of the observed differences. Full article

Dengue is the most prevalent arthropod-borne viral disease, caused by infection with the dengue virus (DENV). Severe dengue is characterized by significant vasculopathy involving a proinflammatory and procoagulant state associated with increased vascular permeability. However, the host–virus interactions driving this process remain incompletely elucidated. Monocytes (Mø) are primary target cells during DENV infection and actively release extracellular vesicles, like microparticles (MPs), mediating intercellular communication, contributing to dengue pathogenesis. Here, we evaluated whether MPs released by DENV-infected monocytes represent a previously underappreciated mechanism contributing to dengue-associated vascular dysfunction. The vascular endothelium plays a determining role in the response to injury because it functions as a regulatory interface during hemostasis (coagulation–fibrinolysis–inflammation) and by preserving the endothelial barrier. We found that these vesicles transport viral proteins (E and NS1), exhibit a procoagulant profile that promotes thrombin generation, and enhance endothelial vascular cell (EVC) activation. DENV-infected THP-1 Mø MPs interaction induces a shift toward a procoagulant, proinflammatory, and proadherent phenotype, characterized by increased expression of PAR-1, TF, ICAM-1, and VCAM-1, reflecting the establishment of a sustained HMEC-1 EVC activation that compromises vascular barrier integrity. This leads to increased permeability, a hallmark of DENV-associated vasculopathy and a central event in the progression to severe dengue. Full article

Compact low-pressure emitters (LPEs) can improve the affordability of drip irrigation, but they must also demonstrate clog resistance for long-term reliability and adoption. Recent research on LPEs has focused on their hydraulic modeling and characterization, but few studies have evaluated or improved their clog resistance. To address this gap, we present a design theory for clog-resistant LPEs and characterize their performance in the lab and field. We focused on the emitters’ weir (or ‘overflow groove’ or ‘channel’), a micrometer-scale internal hydraulic passage, traditionally having a rectangular cross-section. In LPEs, the weir must be shallow to generate the hydraulic resistance required for low-pressure operation, thereby increasing the risk of particulate-jamming-based clogging. A hydraulic model of weirs with arbitrary cross-sections was used to estimate that trapezoidal profiles could be 33–41% deeper than hydraulically equivalent rectangular ones, suggesting that the trade-off between clog resistance and hydraulic performance in LPEs could be navigated through weir cross-section design. To practically validate this proposition, two compact LPEs with trapezoidal weirs (1 and 2 L/h nominal discharge) were designed and tested in the lab and field. Lab results indicated compatibility with 125 $\mathsf{\mu }$m (1 L/h) and 177 $\mathsf{\mu }$m (2 L/h) mesh filters that are typical for these flow rates, providing a basis for field testing the LPEs against commercial emitters. After field tests with these filters, the LPEs held 90–94% of their initial discharge and demonstrated irrigation reliability that was statistically on par with or better than some commercial emitters, despite having 15–65% lower operating pressure. The findings of this work demonstrate the practical viability of compact LPEs for affordable drip irrigation and provide a design framework for their continued development. Full article

Background/Objectives: Urothelial carcinomas are the most common tumors of the bladder. There are limited known cancer stem cell markers in these tumors. Ly6/uPAR gene family members are considered to be markers of cancer stem cells and tissue stem cells in mice, but studies on their expression or role in human cancers are limited. In this study, we aimed to investigate the expression of LY6/uPAR gene family members in human urothelial cancers. Methods: A total of 84 patients were included in the study. Patients diagnosed with urothelial carcinoma were divided into low-grade noninvasive and high-grade invasive carcinoma groups. Normal urothelial samples were used as a control group. RNA isolation was performed from paraffin blocks, and then cDNA was obtained. LY6D, LY6E, LY6H, LY6K, PSCA, LYPD2, SLURP1, GML, GPIHBP1, and LYNX1 genes were analyzed by qRT-PCR method. Results: We observed significantly higher expression of LY6E, LY6K, PSCA, GPIHBP1, and LYNX1 genes in urothelial carcinomas, but lower expression of LY6H, LYPD2, and SLURP1 genes in urothelial cancers compared to the control tissue. Decreased expression of LY6H, PSCA, LYPD2, SLURP1, and GPIHBP1 genes was significantly correlated with poor survival. Conclusions: In the present study, the expression of this gene family in bladder cancer was investigated for the first time in the literature. Given their potential prognostic role and possible relevance as therapeutic targets, this study presents preliminary observations that add to the existing literature. Full article

Manual classification of breast cancer is resource-intensive, slow, and subject to inter-observer variability, motivating automated deep learning solutions. Most current methods rely on unimodal imaging data and struggle with domain generalization (DG) across varied clinical environments. We propose a Dual-Stage Multimodal Alignment approach that integrates breast ultrasound (US) imagery with clinical text reports to improve diagnostic stability. The method proceeds in two stages: (1) Local Correlation Alignment (LCA), which aligns fine-grained visual features with textual embeddings to capture localized lesion attributes, and (2) Global Attention Alignment (GAA), which applies multi-head self-attention to the joint visual–textual sequence to encourage domain-invariant representations. We evaluate the approach on a harmonized, leakage-free repository of 6880 images aggregated from six public US datasets (BUS-CoT, BrEaST, BUS-BRA, BUS-UCLM, BLUI, BUSI) under three protocols: independent benchmarking on BUS-CoT, pooled cross-dataset evaluation, and zero-shot domain generalization on unseen unimodal target domains. On the BUS-CoT benchmark, the 198M-parameter model reaches 0.8177 accuracy and 0.8852 AUC, on par with the 7-billion-parameter Qwen2.5-VL-7B with chain-of-thought reasoning (0.8064 accuracy, 0.8354 AUC) while using roughly 1/35 the parameter count. In the pooled setting, it is competitive with single-domain state-of-the-art methods on individual subsets (e.g., 0.9576 AUC on BUSI, 0.8741 accuracy on BUS-BRA). Under zero-shot transfer without clinical text, per-domain AUC ranges from 0.7360 to 0.8060 across four unseen targets, providing a lower bound under cross-scanner shift. These results indicate that task-specific multimodal alignment can rival large vision-language models in breast US diagnosis at a fraction of the parameter count. Full article

Food waste remains a persistent sustainability challenge within independent restaurants, where operational pressures, cultural norms, and resource constraints limit systematic waste management. This study examines how an industry–academia partnership enabled the co-creation of food-waste reduction practices between hospitality students and sixteen independent restaurant operators in Wellington, New Zealand. Adopting the Participatory Action Research (PAR) approach, data were gathered through semi-structured interviews with operators and focus group discussions with hospitality students. Findings reveal that food wastage in the study units is shaped by time pressure, customer service expectations, tacit kitchen routines, and uncertainty in forecasting the demand. The study identifies three mechanisms—emotional disruption, shared reflection and experimentation—through which sustainability competencies become a part in professional identity. It offers theoretical grounded and practically actionable insights for industry–academia collaboration in resource-constrained hospitality environments. Full article

Background/Objectives: Climate change worry is an emerging concern in youth mental health, but little is known about how it is associated with positive psychological functioning among university students. This study examined whether climate change worry was associated with flourishing and whether this association showed a cross-sectional statistical indirect effect through anxiety and depressive symptoms, with physical activity specified as a first-stage boundary condition. Methods: A cross-sectional anonymous survey was conducted in 2026 using convenience sampling among students from four universities located in three provincial-level regions of China, covering southern, western, and central areas. After predefined quality control procedures, 2826 valid responses were included. Climate change worry, anxiety and depressive symptoms, flourishing, and physical activity were assessed using the Climate Change Worry Scale (CCWS), the Patient Health Questionnaire-4 (PHQ-4), the Flourishing Scale (FS), and the Physical Activity Rating Scale-3 (PARS-3), respectively. Pearson correlations and conditional process analyses were conducted using the PROCESS macro, with 5000 bootstrap samples. Results: Climate change worry was positively associated with anxiety and depressive symptoms (r = 0.331, p < 0.001) and negatively associated with flourishing (r = −0.193, p < 0.001). Anxiety and depressive symptoms were negatively associated with flourishing (r = −0.486, p < 0.001). The cross-sectional statistical indirect effect through anxiety and depressive symptoms was significant (indirect effect = −0.1277, 95% bootstrap CI: [−0.1441, −0.1123]). Physical activity was statistically associated with a weaker first-stage association between climate change worry and anxiety/depressive symptoms (B = −0.0014, p < 0.001; ΔR² = 0.0064). The index of moderated mediation was significant (0.0014, 95% bootstrap CI: [0.0008, 0.0020]). Conclusions: Climate change worry was statistically associated with lower flourishing, primarily through higher anxiety and depressive symptoms. Physical activity was associated with a weaker first-stage association, but the moderation effect was small in practical magnitude. Given the cross-sectional and self-report design, these findings should be interpreted as conditional statistical associations rather than causal or protective effects. Full article

Objectives: To investigate the clinicopathologic characteristics and molecular biomarkers of atypical vasoproliferative retinal tumor (VPRT) with hemangioblastoma-like histopathologic features and concomitant von Willebrand factor (VWF) abnormalities. Methods: A 48-year-old woman undergoing phacoemulsification and 25-gauge pars plana vitrectomy with tumor resection was evaluated. Histopathological findings and immunohistochemical study of the resected tumor were performed using CD34, α-smooth muscle actin (αSMA), and glial fibrillary acidic protein (GFAP) markers. Preoperative plasma and intraoperative vitreous fluid VWF antigen levels, as well as ristocetin cofactor activity, were quantified using latex immunoturbidimetry. Results: Ultra-widefield imaging and angiography demonstrated a peripheral retinal tumor with intense vascular leakage and surrounding capillary nonperfusion. Histopathology showed hyalinized vascular components supportive of VPRT, along with abundant CD34/α-SMA-positive microvessels and scant GFAP-positive glial cells. Notably, numerous foamy vacuolated poorly differentiated cells suggested mixed hemangioblastoma-like features. Preoperative plasma VWF antigen (182.6%) and ristocetin cofactor activity (147.7%) were elevated, and vitreous VWF antigen was successfully detected at a low but distinct level (7.7%).and suggests that VWF abnormalities in the plasma and vitreous may reflect endothelial activation and/or blood–retinal barrier disruption in a subset of vascularized retinal tumors. Conclusions: Our findings demonstrate that VPRT may exhibit mixed clinicopathologic features, including hemangioblastoma-like components, which underscores the necessity of immunohistochemical assessment for definitive diagnosis. Furthermore, the quantification of VWF abnormalities in the plasma and vitreous suggests that VWF serves as a potential biomarker reflecting endothelial activation and/or blood–retinal barrier disruption in vascularized retinal tumors. Full article

Across much of the Arctic, climate warming has reduced the extent of thicker and more persistent sea ice and increased the prevalence of thinner first-year ice. Thin first-year landfast sea ice is ecologically important because reduced ice thickness can increase light transmission to the ice–water interface, while the associated brine conditions, including salinity and permeability, can strongly influence algal biomass accumulation and photophysiology. This thin (0.24–0.55 m), short-lived, seasonal, first-year landfast sea ice already dominates Nuup Kangerlua fjord, southwest Greenland, making it a useful natural example of ice conditions that may become more common in parts of the future Arctic. We focused on late February–early March because this period captures the seasonal transition from very low winter irradiance toward increasing spring light, when sea ice algal communities begin photosynthetic acclimation prior to the main bloom period. Using this site as an example of future Arctic-like conditions, we investigated chlorophyll a (Chl a) concentration and the photobiology of sea ice algal communities during five sampling events between 2017 and 2022. The vertical distribution of Chl a concentration and photobiological parameters measured with variable chlorophyll fluorescence differed between years, as did Chl a concentrations, with integrated biomass ranging from 0.08 to 0.78 mg Chl a m⁻². Direct under-ice PAR measurements showed transmittance values ranging from 0.013 to 0.29. Bottom-ice communities were acclimated to relatively high light intensities, with E_k often exceeding 200 µmol photons m⁻² s⁻¹, and we detected no clear evidence of photoinhibition in the fluorescence data. Boosted regression tree models identified brine salinity as the main predictor of both Chl a concentration, explaining 42.0% of the variation, and, Φ_{PSII_max}, the maximum dark-adapted photosynthetic efficiency, explaining 86.1% of the variation. Both parameters decreased exponentially with increasing sea ice brine salinity (p < 0.0001), indicating that higher brine salinity was associated with reduced algal biomass and lower photosynthetic efficiency. These results show that short-lived first-year landfast sea ice can support physiologically active sea ice algal communities despite relatively low biomass, and suggest that algal performance in this ice type was more strongly associated with brine salinity during the late-winter to early spring sampling period, while light availability also varied substantially among years. As thin and short-lived sea ice conditions become more common in parts of the Arctic, this habitat may represent an increasingly important, though temporally variable, component of Arctic marine primary production. Full article

Marine-derived collagen peptides exhibit potent intestinal barrier protection; however, their gastrointestinal fate and molecular targets remain unclear, limiting their practical applications. This study investigated the digestive stability and transepithelial transport of GPSGPQGSR, a mucoprotective peptide from Alaska pollock (Gadus chalcogrammus) skin, using simulated gastrointestinal digestion, a Caco-2 cell transport model, and an UPLC-QTOF-MS/MS. The results showed that GPSGPQGSR was a digestion-resistant peptide that reached the intestinal epithelium intact. Although brush border membrane enzymes partially hydrolysed the peptide, 42.16% of intact GPSGPQGSR remained in the luminal compartment after 2 h of incubation. No intact peptide was detected in the basolateral compartment. Molecular docking and 100 ns molecular dynamics simulations identified TLR2 (−14.936 kcal/mol) and PAR2 (−10.154 kcal/mol) as high-affinity extracellular targets of GPSGPQGSR, with stable peptide–receptor interactions and extensive hydrogen bonding networks between the peptide and each receptor (RMSD of 1.8 Å and 2.2 Å, respectively). Pharmacological blockade of TLR2 or PAR2 abolished the protective effects of GPSGPQGSR. These findings demonstrate that GPSGPQGSR acts as a digestion-resistant extracellular signalling peptide that reaches the intestinal epithelium intact and protects barrier function through apical TLR2 and PAR2, providing a mechanistic basis for the rational development of marine collagen peptides for improving intestinal health. Full article

The performance of face recognition is highly dependent on detection accuracy and robustness. However, most face recognition methods assume that face regions are perfectly aligned and rely exclusively on visual features within bounding boxes. This assumption is often violated in complex environments, including low resolution, poor illumination, and partial occlusion. To overcome these limitations, this paper proposes a collaborative face detection and recognition framework enhanced by meta-auxiliary learning. The proposed method exploits shared representations by an improved feature extraction module, and introduces gender classification and age estimation as auxiliary tasks to construct a meta-composite loss. Furthermore, facial geometric structure are acquired by extracting facial landmark coordinates from detected face regions. This multi-task architecture imposes implicit constraints on shared features, while simultaneously boosting feature discriminability and system robustness. By fusing facial geometric structures with semantic meta-information, facial identity representations are effectively enriched, leading to substantial performance gains in complex scenarios. Extensive experiments on the SCFace and LRD benchmark datasets validate the effectiveness of the proposed method. On SCFace, our method achieves 77.21% Genuine Acceptance Rate (GAR) at a False Acceptance Rate (FAR) of 1%, surpassing ArcFace, Octuplet Loss, and HDFGD by 1–2 percentage points. On LRD200 and LRD100 under an extreme low-resolution condition of 15 pixels, the recognition accuracies of the proposed method reach 67.98% and 60.98%, respectively—on par with state-of-the-art methods and markedly superior to conventional baselines. These results confirm that meta-auxiliary learning effectively compensates for insufficient visual information by leveraging semantic cues. Non-parametric statistical tests (Wilcoxon signed-rank test) confirm that the improvements achieved by the proposed method are statistically significant (p < 0.05). Full article

To investigate how spreading conditions affect green tea taste and aroma and to develop a generalizable prediction model from small data for process optimization, this study integrated SEM, non-targeted dual-omics, and TabPFN to systematically analyze Echa No. 10 spreading. A central composite design was used. Dehydration-induced mechanical stress altered cell membrane permeability, driving non-volatile taste compound transformation and volatile aroma release. Two chemical-sensory proxies, relative polyphenol-to-amino acid ratio (R-PAR) and floral intensity index (FII), were established using ultra-high performance liquid chromatography–high-resolution mass spectrometry (UHPLC-HRMS) and headspace solid-phase microextraction–gas chromatography–mass spectrometry (HS-SPME-GC-MS). A prediction model was built with these indicators and TabPFN. Multi-objective optimization yielded optimum conditions: initial moisture 76.8%, temperature 26.2 °C, relative humidity 61.5%, air speed 0.85 m/s, achieving R-PAR 0.465 and FII 125.70. Compared with response surface methodology (RSM), partial least squares regression (PLSR), and support vector regression (SVR), TabPFN showed prediction R² of 0.81 and 0.77, showing favorable applicability and predictive capability on small-sample data. This study validates TabPFN’s suitability for small-sample tea processing modeling, quantifies the mapping between spreading and key taste/aroma metabolism, and provides a methodological foundation for digital precision and intelligent optimization in green tea production. Full article

Acinetobacter is one of the most relevant pathogenic and nosocomial bacteria in human medicine. However, in veterinary medicine, particularly in Spain, there are very few studies on the impact and frequency of infections due to this genus. The main objective of this study was to characterise Acinetobacter isolates recovered at the Complutense Veterinary Teaching Hospital of the Complutense University of Madrid (HCV-UCM), with special emphasis on detecting antimicrobial resistance. A total of 23 isolates obtained from different animal species and samples over a 25-year period were included in the study, based on their identification as Acinetobacter by VITEK-2. Identification was made by using MALDI-TOF, VITEK-2, whole-genome sequencing (WGS) and a chromogenic medium. WGS confirmed that 13 of the 23 isolates belonged to Acinetobacter spp. Antimicrobial susceptibility was interpreted according to CLSI guidelines using the Kirby–Bauer disk diffusion and broth microdilution method. The proportion of clinical isolates identified as Acinetobacter spp. at the HCV-UCM was 0.3%. Of these, 15.4% (2/13 isolates) were classified as multidrug-resistant. Two isolates with the highest MIC for tigecycline carried the tet(X) gene, and two isolates harboured mutations in both gyrA and parC QRDR regions. The results of this study suggest that, in this hospital, antimicrobial resistance among Acinetobacter isolates may not yet be widespread. Full article