Search Results (402)

Cross-scene Hyperspectral Image (HSI) classification faces substantial domain shifts caused by sensor heterogeneity, acquisition variation, and scene diversity. While benchmark annotations are assigned to individual center pixels, local patches often contain implicit multi-label semantics due to spectral mixing and spatial overlap. This mismatch distorts prediction structure, exacerbates generalization errors, and limits the effectiveness of standard domain generalization (DG) techniques focused solely on feature or prediction invariance. We propose Class-Aware Semantic Calibration (CASC), a systematic semantic structure calibration framework that addresses three complementary distortions induced by mismatched patch supervision: (i) Balance corrects class frequency bias via reweighted supervision; (ii) Separability enhances boundary decision stability through margin-based logit calibration; and (iii) Independence reduces domain-specific spurious co-occurrence via prediction covariance decorrelation. To preserve calibrated semantics under pseudo-source shift, we further introduce a complementary DualAlign (DA) module, which jointly aligns feature statistics and prediction distributions, enforcing consistency at both representation and semantic levels. Extensive experiments on three cross-scene benchmarks (Houston, Pavia, and WHU-Hi) demonstrate that CASC-DA consistently improves performance over strong baselines, achieving an average gain of 3.0% in overall accuracy and 4.9% in Kappa coefficient compared with the best-performing baseline on each dataset. These results underscore the importance of semantic structure calibration for domain-generalized HSI classification. Full article

This paper develops a boundary-sensitive asymptotic theory for nonparametric conditional U-statistics smoothed by support-adapted asymmetric kernels when the response variable is subject to Missing-at-Random observation. The problem lies at the intersection of three well-established but traditionally separate lines of research: conditional U-statistics, asymmetric smoothing on constrained supports, and incomplete-data inference under MAR sampling. The contribution of the paper is not a novelty claim concerning any of these components in isolation. Rather, it consists in deriving a kernel-specific and MAR-aware limit theory for their simultaneous occurrence, where the estimators are nonlinear complete-case ratios of localized U-statistics and the localization devices are point-dependent approximate identities adapted to the geometry of the covariate support. The analysis covers three principal classes of support-respecting smoothers: Dirichlet kernels on the simplex, Bernstein polynomial smoothers, and multivariate beta kernels on hypercubes, with an additional extension to mixed continuous–categorical regressors. These smoothing schemes are not translation-invariant, and their local moments, effective support, normalizing constants and $L^2$-masses vary with the evaluation point, especially near the boundary. Consequently, their incorporation into conditional U-statistics requires more than a direct transfer of ordinary asymmetric-kernel regression theory. The numerator and denominator of the estimators are localized U-statistics whose stochastic expansions are governed by Hoeffding projections, including canonical components that must be controlled uniformly over the conditioning domain. Under regularity, smoothness and positivity assumptions adapted to the MAR setting, we establish uniform consistency, weak and strong uniform convergence rates, stochastic expansions and asymptotic normality. The results are obtained both on fixed compact subsets and on interior regions approaching the boundary, thereby identifying how support geometry enters the bias and stochastic normalizations. A central feature of the theory is the separation between the deterministic effect of complete-case sampling and its stochastic effect. For the complete-case estimator, the natural deterministic equivalent is obtained by replacing the design density f with the effective complete-case density $pf$, where p is the propensity score. Thus, the MAR mechanism may enter higher-order deterministic bias constants through the local design tilt, whereas the leading stochastic dispersion reflects the loss of effective information through propensity score factors. The precise variance constants and normalizing rates remain kernel-specific, depending on the local $L^2$-structure of the Dirichlet, Bernstein or beta smoothing device. The paper should therefore be viewed as a MAR extension and refinement of the complete-data asymmetric-kernel conditional U-statistic theory. It provides a common probabilistic architecture for several boundary-adapted smoothing schemes while retaining the kernel-dependent bias operators, variance constants, boundary regimes and Hoeffding-projection structures required for sharp asymptotic interpretation. Numerical experiments illustrate the finite-sample behavior predicted by the theory and highlight the interaction between support-adapted smoothing, boundary effects and incomplete response observation. Full article

This work concerns the harmonization of geospatial data to improve linkages between place-based characteristics and health outcomes. Such data are typically available as geographic layers, each representing a distinct attribute (e.g., income or distance to a clinic). Since layers are typically constructed independently, their boundaries tend to be spatially incongruent, which can create inconsistencies and introduce bias. This motivates developing algorithmic approaches for aligning such layers while aiming to preserve spatial integrity. This paper formalizes the problem of aligning k collections of m spatial supports over n spatial units in a d-dimensional Euclidean space such that maximum distortion to any collection is minimized. In the above setting, k is the number of layers; n is an indivisible population unit (e.g., census tract); m denotes supports, which are larger regions aggregating a set of contiguous units in order to capture broader regional patterns or enhance statistical stability; and $d=2$. It is shown that: (1) the one-dimensional case is solvable in time polynomial in k, m, and n; (2) the two-dimensional case is NP-hard for two collections of two supports each; and (3) a heuristic can be provided for aligning a set of collections in the two-dimensional case, which is of practical importance. Full article

Previous research has documented certain associations between physical exercise and sleep quality, yet little is known about the potential influencing mechanisms and boundary conditions underlying them. Thus, the present study aims to examine the potential mediating role of depressive symptoms and the moderating effect of physical health in this relationship. Using the individual-level survey data of 1613 emerging adults aged 18–25 from the China Family Panel Studies (Mage = 20.45 years, SDage = 1.32; 53.8% female), the common method bias test, descriptive statistics and correlation analysis, mediation effect analysis, moderation effect analysis, and simple slope test were sequentially performed using SPSS 27.0, with the significance level set at 5%. The results disclosed that depressive symptoms may play a partial mediating role between physical exercise and sleep quality among emerging adults, and physical health significantly moderated the association between physical exercise and depressive symptoms, indicating a stronger negative association among emerging adults with worse physical health compared to those with better physical health. In addition, exploratory analyses suggested that physical health may also moderate the associations between physical exercise and sleep quality, as well as between depressive symptoms and sleep quality. These findings suggest that emerging adults with lower physical health may often accompany higher depressive symptoms and poorer sleep quality, and also highlight the importance of actively engaging in physical exercise and developing regular exercise habits in daily life to effectively address this problem. Full article

Accurate, non-invasive body weight estimation is essential for management and performance monitoring in beef cattle systems, yet conventional scales and manual measurements require animal handling, infrastructure, and labor. This study presents an integrated pipeline that segments cattle from overhead depth images and predicts body weight from extracted image features. The approach uses a Light Attention Encoder–Decoder (LAED) segmentation model combining depthwise separable convolutions, Gaussian Context Transformer (GCT) attention, a multi-scale dilated bottleneck, and dual heads for region and boundary prediction. Depth videos were collected using an overhead Intel RealSense D435 RGB-D camera from 60 beef heifers. To reduce animal-level leakage, leave-one-animal-out cross-validation was used for segmentation. LAED + GCT achieved 96.91% Dice (95% confidence interval (CI): 96.56–97.21%) and 94.22% IoU (95% CI: 93.58–94.77%), while operating at 33.08 frames per second. For weight prediction, biometric traits and deep features were evaluated using random forest, support vector regression, and fully connected neural networks. The best primary-metric body-weight model used biometric traits with support vector regression, achieving MAPE = 6.75%, pooled $R^2$ = 0.68, MAE = 23.92 kg, and RMSE = 31.79 kg. Among FCNN models trained independently within each cattle-level fold, the best result used ResNet50 features and achieved MAPE = 7.76%, a pooled $R^2$ = 0.56, an MAE = 27.60 kg, and an RMSE = 37.07 kg. The mean signed prediction bias for the biometric-SVR model was −1.04 kg, using predicted minus observed body weight, with a bootstrap 95% confidence interval of −9.63 to 7.41 kg. These results support the promise of overhead depth imaging for non-invasive cattle body segmentation and weight estimation, while larger external validation remains necessary. Full article

Background: Brain metastases are the most common intracranial tumors in adults and are traditionally considered well-demarcated lesions amenable to complete surgical resection. Nonetheless, increasing histopathological evidence demonstrates that metastatic cells may infiltrate beyond the contrast-enhancing margin into surrounding brain parenchyma, challenging the reliability of conventional imaging for defining true tumor boundaries. Confocal laser endomicroscopy (CLE) using Sodium Fluorescein (SF) has emerged as a novel intraoperative imaging modality capable of providing real-time, high-resolution optical biopsies, potentially improving margin assessment during metastasis surgery. Methods: A systematic literature search was performed according to PRISMA guidelines across PubMed, Embase, Scopus, Cochrane Library, and Google Scholar up to 3 March 2026. Studies evaluating intraoperative CLE with SF in adult patients with brain metastases were included. Data regarding study design, patient population, CLE system, imaging characteristics, and diagnostic performance were extracted. Risk of bias was assessed using the QUADAS-2 tool. Results: Ten studies met the inclusion criteria for qualitative synthesis, comprising over 650 patients; however, most studies included heterogeneous intracranial tumor populations, with only a subset specifically involving brain metastases. CLE enabled real-time visualization of tumor microarchitecture and demonstrated high sensitivity for tumor detection, frequently exceeding 90% in prospective studies. Specificity varied across studies, reflecting challenges in distinguishing tumor infiltration from reactive tissue at the tumor–brain interface. The MetInfilt trial highlighted that infiltrative growth patterns are common in brain metastases and can be visualized intraoperatively using CLE. Additional studies demonstrated that fluorescein-based CLE allows differentiation of tumor zones and may facilitate targeted margin assessment; however, evidence demonstrating improvement in clinically meaningful outcomes such as extent of resection, local recurrence, progression-free survival, or overall survival remains limited. Conclusions: Confocal laser endomicroscopy using SF represents a promising intraoperative adjunct for assessing tumor margins in brain metastasis surgery. By enabling real-time microscopic visualization of the metastasis–brain interface, CLE may support a more biologically informed surgical strategy. Full article

Although small and medium enterprises (SMEs) anchor employment and output across Sub-Saharan Africa, their uptake of artificial intelligence (AI) lags global benchmarks, and prevailing explanations dwell on capital, infrastructure, and institutional voids while overlooking the leadership competencies that determine whether available resources are mobilised at all. Addressing this gap, the present study asks how the digital leadership capabilities of SME owner-managers shape their intention to adopt AI in Nigeria, and through what organisational mechanisms and under what boundary conditions this influence operates. Anchored in the Diffusion of Innovation Theory and the Tigre–Henriques–Curado model of digital leadership, a cross-sectional survey was administered to owner-managers of registered SMEs drawn from six states; a sample of 390 was derived from a population of 23,290 firms using the Taro Yamane formula with proportionate allocation, and 306 valid responses were retained. Partial Least Squares Structural Equation Modelling (WarpPLS 8.0) was applied after confirming reliability (Cronbach’s $\alpha$: 0.69–0.84; composite reliability: 0.83–0.88), convergent validity (AVE: 0.56–0.67), and common method bias control. Strategic ($\beta$ = 0.298), interpersonal ($\beta$ = 0.245), and personal attribute ($\beta$ = 0.129) capabilities each significantly raised AI adoption intention. In contrast, delivery-related capabilities ($\beta$ = 0.090, p = 0.057) did not, indicating that pre-adoption intention is governed by cognitive-strategic and relational competencies rather than execution skills. Organisational innovation climate partially transmitted the effects of strategic and interpersonal capabilities, and firm size amplified the interpersonal pathway in medium-sized firms. The study contributes a leadership-centred account of AI adoption in an under-researched African setting and, by estimating mediation and moderation within a single framework, clarifies both why and when digital leadership translates into AI readiness, yielding capability-specific guidance for owner-managers and SME support policy. Full article

This study examines equifinality and compensatory calibration in hydrodynamic modelling of semi-enclosed coastal systems, using the Xiamen–Kinmen coastal waters as a representative tide-dominated case. A progressive diagnostic framework based on the normalized marginal contribution rate (MCR) was developed to quantify the relative effects of open-boundary forcing, spatially heterogeneous bottom friction, and atmospheric forcing within a depth-averaged barotropic model. Multi-metric validation against in situ water-level and depth-averaged current observations shows that the physical consistency of open-boundary forcing is the dominant control on model skill, particularly in reducing systematic elevation bias within the embayment. Bottom-friction parameterization produces more localized and site-dependent improvements, mainly affecting the spatial structure of current speed and direction under geomorphological constraints. Atmospheric forcing contributes only limited marginal gains during the study period, with modest directional corrections under weaker tidal conditions. These results indicate that hydrodynamic optimization for semi-enclosed bays should prioritize boundary consistency before local parameter tuning, thereby reducing compensatory calibration risk and improving physical interpretability. Remaining localized velocity errors in estuaries and high-curvature channels highlight the limitations of the depth-averaged barotropic assumption, under which density-driven baroclinic flows and vertical secondary circulations cannot be explicitly resolved. The proposed framework provides a reproducible approach for diagnosing and optimizing nearshore hydrodynamic models. Full article

Landslide detection has largely relied on supervised learning or DEM-based representations, which can limit rapid deployment and generalization across heterogeneous terrain. In this work, we present a zero-shot, fully unsupervised framework that identifies landslide-like geomorphic instability candidates from raw UAV-mounted LiDAR, removing the need for labeled data, pre-event baselines, or rasterized terrain abstractions. Our approach is motivated by the observation that landslides manifest as localized geometric inconsistencies in the terrain surface. We capture this through a multi-scale formulation that combines point-level and cluster-level indicators of instability. At the point level, a PCA-based residual depth metric reduces slope-induced bias and highlights surface discontinuities, while local concavity captures terrain depletion patterns. At the cluster level, geomorphometric descriptors such as curvature concentration, surface roughness, elevation discontinuity, and slope variation are extracted using density-aware 3D clustering and integrated through adaptive feature fusion. The resulting probabilistic instability field enables spatially coherent delineation of landslide scars, including rupture boundaries, displaced material, and emerging failure regions. In addition, the detected patches provide useful priors for post-event susceptibility analysis without requiring temporal observations. Experiments across diverse geomorphic settings show that the proposed method improves detection of subtle terrain disturbances compared to DEM-based pipelines and supervised learning approaches, while remaining robust to noise and terrain variability. Overall, this work demonstrates that geometry-driven, unsupervised inference on raw 3D data can serve as a practical and scalable alternative for near real-time landslide detection using UAV-based systems. Full article

Accurate segmentation of ship targets in high-resolution remote sensing images is crucial for maritime monitoring, traffic management and naval security. However, existing methods struggle to simultaneously address extreme scale variations in ships and severe complex background interference, leading to unsatisfactory accuracy and generalization in scenarios with shoreline occlusion and ocean wave noise. To tackle this challenge, we first construct a large-scale, high-quality multi-scale ship dataset containing 69,407 professionally annotated samples. Then, we propose ShipMS-BSNet, a multi-scale feature fusion network based on nnU-Net. At the encoder, the Multi-Scale Receptive Field Enhancement (MSRF) module captures multi-scale contextual information, while the Background Suppression Channel Attention (BSCA) module suppresses invalid background responses via learnable negative bias. At the decoder, dynamic upsampling restores spatial details, and a final Multi-Scale Refinement (MSR) module optimizes target boundaries. Extensive experiments on our self-built dataset and the public HRSC2016 dataset show that our method outperforms mainstream approaches. On the self-built dataset, it achieves 0.879 precision, 0.875 Recall, 0.868 F1-score and 0.761 IoU, validating its strong robustness for multi-scale ship segmentation in complex marine environments. Full article

Large language models (LLMs) are increasingly used to support writing, reasoning, translation, and decision-making, often on the assumption that easier access to information improves judgment. This integrative conceptual review argues that this assumption is incomplete because LLMs interact not with neutral information processors, but with users who bring prior beliefs, directional motivations, cognitive-effort constraints, and varying willingness to verify. The article develops the concept of artificial confidence: a relational and systemically reinforced form of unwarranted certainty that emerges when prompt-shaped, fluent, and seemingly authoritative AI outputs are experienced as independent validation. Drawing on research in judgment and decision-making, motivated reasoning, automation bias, processing fluency, epistemic vigilance, LLM sycophancy, and systems thinking, the review distinguishes artificial confidence from related constructs and proposes a socio-technical feedback model linking user motivations, prompt framing, model accommodation, perceived validation, reduced verification, and institutional normalization. The framework also identifies boundary conditions under which LLMs can improve judgment by preserving epistemic friction, source checking, uncertainty awareness, and accountability. The article concludes by offering operational definitions, behavioral indicators, testable hypotheses, and design and governance implications for AI-augmented systems in which human judgment remains revisable, accountable, and evidence-sensitive. Full article

Silicon resonant accelerometers generate sinusoidal outputs with frequency shifts that carry acceleration information. At high update rates, conventional counting-based readout suffers from gate-boundary timing quantization. This work proposes a high-update-rate frequency readout method that reconstructs frequency from the continuous phase evolution of the original sinusoidal resonant signal through quadrature demodulation, phase extraction, and phase difference rather than waveform reshaping and edge counting. To implement the proposed readout chain, an FLL–PLL cooperative loop was included to assist coarse acquisition and fine tracking on a Zynq-7020 platform. This study focuses on the readout principle, FPGA implementation, and prototype-level evaluation. At a 1 kHz update rate, the proposed method showed a lower theoretical quantization limit than the synchronous multi-cycle counting method. Under room-temperature conditions, after a 30 min startup, the proposed method reduced the standard deviation of the 1-second-averaged zero-bias output over 1800–5400 s from 4.1 μg to 2.4 μg and reduced the frequency-difference peak-to-peak value from 0.03743 Hz to 0.02410 Hz. These results support the feasibility and practical value of the proposed method for high-update-rate readout of sinusoidal resonant signals under the tested steady-state conditions. Full article

Accurate classification of transformer partial discharge (PD) severity is essential for insulation diagnostics yet remains challenging due to nonlinear feature relationships and class imbalance. This study evaluates whether feature extraction improves PD severity classification and compares the effectiveness of linear and nonlinear extraction methods. A dataset of 294 samples was categorized into four IEC-aligned severity classes. Two raw measurements (discharge magnitude and applied voltage) were expanded into a 15-dimensional feature space. Principal Component Analysis (PCA) and a bottleneck Autoencoder (AE) were used for linear and nonlinear feature extraction, respectively. Extracted features were classified using an identical Multilayer Perceptron (MLP). Both feature extraction methods improved classification performance over raw and full-feature baselines (96.6%). PCA+ANN achieved 100.0% accuracy (k = 9), while AE+ANN achieved 98.3% (k = 8). The AE misclassified one minority “Normal” sample due to poor latent boundary representation. Reconstruction analysis showed the highest error for the Normal class, reflecting imbalance-driven optimization bias. Feature extraction enhances PD severity classification, with linear PCA outperforming nonlinear AE in this near-linearly separable dataset. PCA’s deterministic projection preserves minority class boundaries more effectively, whereas AE performance is limited by class imbalance. These findings suggest that nonlinear methods provide advantages only in more complex feature spaces. Full article

This paper applies physics-informed neural networks (PINNs) to laterally excited liquid sloshing in a two-dimensional rectangular tank, where near-resonant forcing (${\omega }_e/{\omega }_1=0.9$) produces a multi-frequency beating response with a period of approximately $10T_1$. Linearized potential flow theory governs the problem; the network learns the velocity potential $\phi (x,z,t)$ while the free-surface elevation $\eta$ is injected analytically. Two training obstacles specific to forced sloshing are analyzed. First, a zero-solution trap arises because the trivial solution $\widehat{\phi }=0$ satisfies all equations except the free-surface conditions, whose residuals are roughly ${10}^{-4}$ times smaller than the Laplace residual; characteristic-scale normalization combined with loss weighting (${\lambda }_D={\lambda }_K=100$) breaks this trap. Second, spectral bias prevents standard MLPs from resolving the three co-existing frequencies (${\omega }_1$, ${\omega }_e$, $\mathsf{\Delta }\omega$); a Fourier time embedding that augments the input from 3 to 9 dimensions overcomes this limitation. Two additional techniques further reduce errors: a hard-wall boundary condition enforced exactly via a $\cos (\pi x/B)$ spatial embedding, which eliminates wall collocation points; and a gradient-enhanced Laplace regularizer (${\parallel \nabla \left({\nabla }^2\widehat{\phi }\right)\parallel }^2$) that constrains velocity smoothness through third-order automatic differentiation. An ablation study shows that these four techniques progressively reduce the horizontal velocity error from ${\epsilon }_u=12.46\%$ to $0.84\%$. Results are validated against a viscous finite-difference benchmark. Over one beating cycle the errors are ${\epsilon }_{\eta }=0.15\%$, ${\epsilon }_u=0.84\%$, and ${\epsilon }_w=1.65\%$. A frequency parameter study across ${\omega }_e/{\omega }_1$ = 0.5–1.1 gives ${\epsilon }_{\eta }<0.25\%$ and ${\epsilon }_u<2.3\%$ for all near-resonance cases. For long-time simulation, a time-domain decomposition strategy with transfer learning partitions the domain into one-beat windows; extending to five beating cycles ($50T_1$) yields ${\epsilon }_u=3.43\%$ and ${\epsilon }_{\eta }=0.30\%$ with no monotonic error accumulation across windows. The methodology is then extended to a three-dimensional rectangular tank ($B\times W\times H$) with bi-directional lateral excitation. The 3-D formulation introduces the y-dimension into the Laplace equation (${\nabla }^2\phi ={\phi }_{xx}+{\phi }_{yy}+{\phi }_{zz}=0$), adds transverse wall boundary conditions ($\partial \phi /\partial y=0$) enforced exactly via a $\cos (\pi y/W)$ embedding, and extends the Fourier time embedding from 9 to 16 dimensions to accommodate six physical frequencies. The bi-directional excitation excites both $(m,0)$ and $(0,n)$ modal families, producing a genuinely three-dimensional beating response. Experimental results verify that the proposed methods can be well generalized to three-dimensional scenarios. Within a single beating cycle, the relative errors reach ${\epsilon }_{\eta }=0.24\%$, ${\epsilon }_u=1.31\%$, ${\epsilon }_v=1.78\%$ and ${\epsilon }_w=2.32\%$, with a total training time of 2499 s. By applying time domain decomposition to carry out two-cycle three-dimensional simulations, the model can steadily maintain satisfactory prediction precision across segmented time intervals, achieving overall errors of ${\epsilon }_{\eta }=0.30\%$ and ${\epsilon }_u=1.32\%$. Full article

The rapid expansion of the Ubiquitous Electric Internet of Things (UEIoT) has introduced a vast array of heterogeneous devices into smart grids, rendering traditional identification methods inadequate. The continuous emergence of new terminal models and frequent firmware updates create a dynamic environment where training data cannot realistically cover all evolving device types. To bridge this gap, we propose HALO (Hierarchical Attribute-guided Learning with Offset Calibration), a generalized zero-shot learning (GZSL) framework specifically designed for IoT device identification. First, a lightweight Transformer-based architecture, NetFormer, is utilized to extract discriminative features by capturing fine-grained temporal behaviors with minimal computational overhead. Second, a Weighted Conditional Variational Autoencoder (W-CVAE) is developed to synthesize high-quality pseudo-samples for unseen classes. To ensure semantic fidelity, the W-CVAE incorporates multi-scale Maximum Mean Discrepancy (MMD) to prevent mode collapse and employs attribute-feature contrastive learning to align semantic and feature spaces. Finally, a hybrid prototype construction strategy and an adaptive bias calibration mechanism are introduced to dynamically adjust decision boundaries, effectively mitigating the seen-class bias inherent in GZSL. Experimental results demonstrate that HALO significantly outperforms existing baseline methods across multiple evaluation metrics, validating the effectiveness and superiority of the proposed framework. Full article