Search Results (847)

This research advances the study of learning under non-stationary conditions by proposing a unified framework for concept drift detection and adaptive regression in evolving data streams. Unlike traditional batch models that assume static data distributions, the proposed approach operates sequentially, enabling real-time adaptation to drifting concepts in both time series and regression tasks. The method integrates Least Squares Support Vector Regression (LS-SVR) with Least Squares Support Vector Data Description (LS-SVDD) to jointly perform prediction and drift monitoring within a single kernel-based structure. LS-SVDD serves as a distributional drift detector, while LS-SVR incrementally updates model parameters to maintain predictive accuracy as data evolves. The framework accommodates both abrupt and gradual drifts, making it suitable for dynamic, high-dimensional environments. Experimental evaluations on synthetic data show that this proposal is able to outperform conventional batch and static methods in accuracy, responsiveness and computational efficiency. This method was compared using a real-world dataset, namely the high-dimensional Drosophila microarray time series, to demonstrate that the proposed approach is able to detect the meaningful change points using the whole data which is not doable using existing methods. Existing methods only used subsets of the dataset. These results highlight the potential of LS-SVR and LS-SVDD integration for real-time, adaptive learning across diverse domains where data distributions change over time. Full article

This paper develops a unified asymptotic theory for conditional single-index U-statistics and the associated conditional U-processes in the setting of locally stationary functional time series subject to missing-at-random response mechanisms. The proposed framework addresses, within a single nonparametric inferential architecture, three major sources of complexity in modern functional data analysis: infinite-dimensional covariates, smoothly time-varying stochastic dynamics, and incomplete response observations. The methodology is based on a class of kernel-type estimators combining temporal localization, functional single-index smoothing, and inverse-propensity correction. Temporal localization captures the gradual evolution of the underlying regression structure, the single-index projection provides an effective dimension-reduction mechanism for functional covariates, and the propensity adjustment restores the target conditional functional under the MAR sampling scheme. The principal contribution of the paper is the establishment of weak convergence, in a suitable space of bounded functions, for the resulting propensity-adjusted conditional U-process indexed by a general class of measurable kernels. Under absolute regularity conditions, local stationarity assumptions, small-ball probability requirements, entropy restrictions of VC type, and uniform consistency of the propensity-score estimator, the normalized process is shown to converge weakly to a tight centered Gaussian process. The limiting covariance structure explicitly reflects the interaction between temporal smoothing, functional concentration, dependence, and the random loss of responses. In parallel, uniform convergence rates are derived for the associated conditional single-index U-statistic estimators, thereby quantifying the respective contributions of smoothing bias, stochastic fluctuation, local-stationarity approximation error, and missingness-induced variance inflation. A substantial part of the analysis is devoted to the technical difficulties created by the simultaneous presence of dependence, nonstationarity, functional covariates, and incomplete observations. The proofs combine Hoeffding-type decompositions adapted to weighted incomplete data, blocking and coupling arguments for absolutely regular triangular arrays, refined entropy bounds for kernel-indexed function classes, and small-ball probability techniques for functional covariates. The MAR mechanism is incorporated via inverse-propensity weighting, and its effects on the effective sample size, asymptotic variance, and bias structure are made explicit. The theory also provides a rigorous foundation for bandwidth selection through blocked, propensity-adjusted cross-validation and clarifies its relation to the corresponding oracle risk. The proposed framework encompasses a broad class of statistical learning and inference problems involving pairwise or higher-order functionals of functional time series. In particular, it applies to conditional Kendall-type functionals, discrimination problems, metric learning with incomplete labels, and conditional independence testing under local stationarity. A simulation study illustrates the finite-sample behavior of the proposed estimators and supports the theoretical findings across varying regimes of temporal nonstationarity, serial dependence, functional concentration, and response missingness. Overall, the results provide a mathematically rigorous and methodologically flexible foundation for inference from evolving functional data when dependence, infinite dimensionality, and incomplete observation are present simultaneously. Full article

Accurately obtaining aircraft aerodynamic parameters is essential for improving flight performance, optimizing design and control strategies, and ensuring flight safety. In this study, the improved Northern Goshawk Optimization (SPNGO) algorithm is used to optimize the kernel parameters and regularization coefficients of the Kernel Extreme Learning Machine (KELM). To address the defects of the original NGO algorithm, such as insufficient global optimization ability and being prone to falling into local optimums, two improvement strategies are proposed. The enhanced SPNGO algorithm is verified by 14 benchmark test functions, and the proposed SPNGO-KELM model is evaluated using open-source F-16 nonlinear simulation data for longitudinal aerodynamic parameter identification. The results demonstrate its effectiveness under the considered simulation conditions, while further validation with real flight-test data is required before application to actual flight environments. Comparative analysis with KELM, NGO-KELM, SSA-KELM, and WOA-KELM models shows that a single KELM is difficult to achieve high-precision aerodynamic parameter identification, and other comparison models have obvious fitting deviations in non-steady-state and strong nonlinear regions. Notably, the SPNGO-KELM model achieves the best identification performance, with a determination coefficient (R²) of 0.96537 and a mean absolute percentage error (MAPE) as low as 3.1574%. Its comprehensive identification accuracy is 1.81% to 37.98% higher than that of the comparison models, and it can effectively suppress error oscillations in nonlinear regions. Experimental results show that the proposed algorithm has excellent identification accuracy, generalization ability, and anti-interference performance. Full article

Reliable prediction of the factor of safety (Fs) of circular-failure soil slopes is critical to geotechnical practice. Data-driven models developed on small slope-stability datasets are, however, prone to overfitting, data leakage, and optimistic bias, which can lead to overestimated predictive performance. This study presents a small-sample-oriented, leakage-aware support vector regression (SVR) framework with a radial basis function (RBF) kernel for Fs prediction. A database of 80 published circular-failure slope cases was compiled, and six predictors were adopted: soil unit weight, slope height, pore pressure ratio, cohesion, internal friction angle, and slope angle. To improve reliability under limited-data conditions, preprocessing, hyperparameter tuning, and performance evaluation were all embedded within a repeated nested cross-validation framework. The proposed SVR model was benchmarked against the back-propagation neural network (BPNN) and radial basis function neural network (RBFNN) models under identical validation partitions and evaluation settings. The results indicated that SVR achieved the best predictive performance among the three candidate models. For case-level illustration, a single representative hold-out split was reported in addition to the repeated nested cross-validation results, on which the SVR model attained an R² of 86.56%, an RMSE of 0.07497, an MAE of 0.0666, and an MRE of 5.29%. In this test subset, all SVR predictions exhibited relative errors below 10%, indicating more stable predictive behaviour than the benchmark models. The main contribution of this study is thus a validated SVR framework for small-sample conditions. Full article

Fully integrated MITC4 (mixed interpolation of tensorial components) shells remain costly for large-deformation sheet-metal forming benchmarks at production mesh sizes. This paper presents a GPU-resident explicit MITC4 shell solver, implemented as a single CUDA pipeline in which co-rotational kinematics, assumed natural strain transverse shear, through-thickness J2 elasto-plasticity, and rigid-surface penalty contact remain in device memory. The study is positioned as computational verification and benchmarking for the Nakajima hemispherical-dome forming benchmark. Canonical shell tests verify the element kernel through membrane and bending patches and a force-driven cantilever, with the cantilever deflection agreeing with the MacNeal–Harder reference within about 2%. On the 10K-element Nakajima benchmark, the present solver agrees with Abaqus/Explicit with a mean von Mises error of 2.95% over 94% of specimen elements and a maximum shell thickness error of 2.08%. In the clamped/binder transition band, section-mean von Mises agrees to +1.0%, whereas section-maximum stress is under-predicted by 10.9%. A 50K-element Abaqus check remains bounded at 80 mm stroke, with section-mean von Mises differences of +0.6% globally and +0.4% in the transition band. For throughput, a separate 500K-element deck over 1.0 × 10⁻³ s and 15,808 steps give per-step speed-ups of 43.7×, 17.7×, and 13.5× versus 1-, 8-, and 32-core LS-DYNA MPP. Full article

Engineered cementitious composite (ECC) is a high-performance strain-hardening material widely used in durable infrastructure, yet its complex multi-parameter interactions make accurate mixture design and performance prediction challenging. This study aims to establish an EDFrame, which is an integrated prediction framework for engineered cementitious composite (ECC). First, two original datasets of ECC’s tensile stress and strain are collected from the comprehensive and authoritative literature, comprising 18 features and 10 categories of single or hybrid fibers. Data augmentation is then performed using a constraints-modified Conditional Tabular Generative Adversarial Network (Tuned-CTGAN), with two traditional methods for comparison. A One-Dimensional Convolutional Neural Network with a residual module (1D-Residual CNN) is developed to predict tensile stress and strain, and its performance was compared against five popular machine learning models. The interpretability of the proposed model has been achieved through Partial Dependence Plot (PDP) and Kernel SHAP analyses. The results demonstrate that Tuned-CTGAN effectively generates reliable synthetic data, significantly improving the R² of 1D-Residual CNN from 0.8658 to 0.9128 for tensile stress and from 0.8433 to 0.9378 for tensile strain, outperforming all compared models. PDP analysis identifies optimal fiber content (1.5–2%) and fiber length (12–20 mm) ranges for enhanced tensile performance, while SHAP analysis reveals fiber length and diameter as the most critical features influencing tensile stress and strain, respectively. The proposed EDFrame provides a robust and interpretable solution for ECC performance prediction, supporting efficient and accurate mixture design in engineering practice. Full article

Background: Heavy metals and per- and polyfluoroalkyl substances (PFAS) are widespread environmental pollutants that have been linked to worsening cognition, but how the two classes act together to shape cognitive function is still not well characterized. Drawing on data from the National Health and Nutrition Examination Survey (NHANES), this observational analysis evaluated how PFAS and metals are jointly related to performance across distinct cognitive domains in older adults. Methods: We analyzed 1447 adults aged 60 years and older from the 2011–2012 NHANES cycle in a cross-sectional design study. Metal levels in serum and whole blood were determined with standardized laboratory assays. Associations of single exposures and of the overall mixture with the CERAD word-learning and recall tasks, Animal Fluency, and the Digit Symbol Substitution Test were assessed using multivariable linear regression, together with Bayesian Kernel Machine Regression (BKMR). Results: Single-exposure models produced largely modest and inconsistent associations across the cognitive measures. Within the mixture models, PFAS, especially PFOA, PFDE, and PFOS, were repeatedly flagged as influential across several domains, whereas the metals tended to matter for specific outcomes only. The strongest negative signals at elevated joint exposure emerged for memory-related measures, notably CERAD Trials 1 and 2. Conclusions: Joint exposure to PFAS and heavy metals appears to influence cognitive domains unevenly, with memory-related measures seeming more responsive as combined exposure rises. These results reinforce the value of mixture-oriented analytic strategies when investigating environmental contaminants in relation to cognitive aging. Full article

Single image dehazing remains a fundamental challenge in computer vision due to the ill-posed nature of the inverse problem and the spatial heterogeneity of real atmospheric haze. Existing convolutional approaches suffer from two structural deficiencies: bounded receptive fields that fail to model large-scale haze gradients, and isotropic kernels insensitive to the directional patterns of atmospheric scattering. This paper proposes ACTA-AOD, a lightweight end-to-end dehazing network that addresses both limitations within a unified framework built upon the AOD-Net K-parameterization. The network integrates two complementary modules: (1) W-EMSAv2, a windowed efficient multi-scale attention module that reduces attention complexity from $\mathcal{O}(N^{2}C)$ to $\mathcal{O}(N{M}^{2}C/4)$ while preserving full-spectrum spatial information through pixel-shuffle reconstruction; and (2) the ACTA Fusion module, which combines structural-reparameterization-based asymmetric convolution with cross-dimensional Triple Attention for direction-sensitive local detail recovery at zero inference-time overhead. On the RESIDE benchmark, ACTA-AOD achieves peak signal-to-noise ratio (PSNR) of 26.02 dB and structural similarity index measure (SSIM) of 0.910 on indoor synthetic data, and 26.13 dB/0.910 on outdoor synthetic data, surpassing the AOD-Net baseline by $+3.41$ dB (indoor) and $+3.58$ dB (outdoor) in PSNR, and exceeding the strongest learning-based baseline (AECRNet, CVPR 2021) by $+1.17$ dB (indoor) and $+1.75$ dB (outdoor). The model processes images at 81 frames per second on a single GPU. Ablation studies and stratified robustness evaluation across five haze density levels confirm the complementary, synergistic contribution of each module. Full article

As a key driver of sustainable urban development, the digital economy transforms urban spatial structures through novel organizational forms such as digital enterprises. Understanding the spatiotemporal distribution of these enterprises is crucial for fostering equitable and efficient urban growth. Focusing on Hangzhou, a leading digital city in China, this study applies kernel density estimation, the standard deviational ellipse, and the nearest neighbor index to analyze the evolution patterns of newly established digital enterprises (NDEs) from 2010 to 2020. It further integrates geodetector and multiscale geographically weighted regression (MGWR) to uncover the drivers behind their spatial differentiation. The results indicate that: (1) The spatial pattern of NDEs evolved from “single-core diffusion” to a “dual-core with multi-center and axial contiguous” structure, yet the density gap between cores and peripheral counties persisted. (2) NDEs exhibited increasing spatial agglomeration over time. (3) Global drivers: the nighttime light index exerts the strongest positive effect, while land costs and population density show negative effects, reflecting cost-squeeze and decentralized locational preferences. (4) Locally, bus accessibility, innovation level and science-education-culture level, display strong spatial heterogeneity; innovation level has very high positive coefficients in innovation poles but negative effects in ecologically sensitive or deindustrialized areas, revealing an “innovation multiplier effect” alongside resource misallocation risks. These findings provide empirical evidence of how digital economy actors spatially manifest, offering insights for urban planners and policymakers to leverage digital growth for guiding sustainable spatial restructuring, enhancing resource allocation efficiency, and promoting balanced regional development. Full article

We demonstrate a loop-based deep optical convolutional neural network that reuses a single free-space optical hardware to realize network depth through repeated passes. Convolution is implemented with programmable SLM with Fourier plane kernels, nonlinearity is provided by the photorefractive phase-only response of a BSO crystal and converted to an effective intensity activation via spatial filtering, and pooling is performed optically using demagnified imaging with an iris. On MNIST, the BSO-based nonlinearity improves test accuracy from 90.8% (linear) to 95.7%, with optimal operation. We model realistic optical noises (laser fluctuation, aberration, detector misalignment, and dust) and compare them using an SSIM-normalized severity metric. Under noise at (s = 0.35) on Fashion-MNIST, accuracy drops from 88.53% (clean) to 79.5% (noisy inference); a feature-level noise-aware training strategy recovers performance to 86.87%. Together, these advances demonstrate that a compact, loop-based hybrid DOCNN, completed with simple optical nonlinearities, simplified pooling, and noise-aware learning, can improve accuracy under realistic conditions. Full article

Rolling bearing fault diagnosis is critical for ensuring the safe operation of rotating machinery, yet it faces significant challenges in noisy environments. This paper proposes BEAM-Net (Bearing-spectrum Enhanced by EMA and Weighted Spectral Convolution Network), a lightweight neural network designed specifically for rolling bearing fault diagnosis under strong noise conditions. Classifying bearing faults from vibration signals remains a challenging task when fault-related features are subtle and easily submerged in background noise—especially when the signal-to-noise ratio (SNR) is low. To address this challenge, BEAM-Net adopts a “decompose–enhance–extract” pipeline: first, an Exponential-Moving-Average Trend Decomposer (ETD) splits the frequency spectrum into a smooth trend component and a fault-sensitive residual component; second, a Spectral Residual Gate (SRG) reinjects detailed residual information through a learnable gating mechanism; finally, a Weighted Spectrum Convolution block (WSC) incorporates a symmetric center-emphasizing prior into the convolution kernel, ensuring that local spectral patterns receive greater attention. Experimental results on the Case Western Reserve University (CWRU) bearing dataset at SNR = −6 dB show that BEAM-Net achieves an F1 score of 99.15% with only 2835 parameters. Compared to the single-convolution baseline, this represents a $+0.78\%$ improvement in F1 score and a 50% reduction in the false positive rate (from $0.18\%$ to $0.09\%$). Cross-dataset validation on the Paderborn University (PU) and Machinery Failure Prevention Technology (MFPT) datasets further confirms the generalizability of the proposed approach, achieving F1 scores of 97.83% and 98.46%, respectively, under comparable noise conditions. These findings demonstrate that combining explicit spectral trend modeling with weighted convolution is not only effective but also parameter-efficient, making it well-suited for noise-robust rolling bearing fault diagnosis. It should be noted that the current method is primarily validated on spectral-analysis-based diagnostics of rolling bearings; its applicability to other vibroacoustic diagnostic modalities (e.g., tapping or nonlinear vibration excitation) and to quantitative defect severity grading remains to be investigated in future work. Full article

Traditional villages in the Chaoshan region serve as living repositories of local cultural heritage. Their concentrated and coordinated conservation and utilization can transcend administrative boundaries, enabling the integrated allocation of regional resources and the enhancement of cultural synergy. Currently, conservation practices for traditional villages are largely limited to piecemeal rescue efforts focused on individual villages. There is a lack of systematic understanding from a regional perspective and an explanation of the mechanisms underlying the formation of local landscapes, which hinders the realization of economies of scale in conservation and the development of cultural synergy. To explore effective approaches for the cluster-based conservation of traditional villages in China’s Lingnan coastal region, as well as the characteristics of human–land relationships and their adaptive mechanisms, this study focuses on 115 national and provincial-level traditional villages in the Chaoshan region. By introducing methods of single-factor and multi-factor cluster identification, the study innovatively constructs a four-dimensional cluster identification framework comprising “spatial proximity, geomorphological similarity, cultural convergence, and residential isomorphism,” and, utilizing the ArcGIS platform for coupled analysis, kernel density analysis, cluster identification, and field surveys, systematically analyzed the diverse typologies and landscape-specific characteristics of traditional village clusters in the Chaoshan region. The results indicate the following: (1) The identification of Chaozhou–Shantou traditional village clusters reveals three diverse types—comprehensive, distinctive, and potential—reflecting the richness and diversity of these clusters in the region. (2) Spatially proximate clusters exhibit a single-core, multi-point distribution, topographically similar clusters show differentiated distributions across plains and river valleys, culturally convergent clusters are significantly correlated with cultural carriers such as postal routes, water transport, and trade, and residential distributions are significantly correlated with topography and landforms, collectively constituting the unique character of Chaozhou–Shantou traditional village clusters. (3) Traditional villages in Chaoshan exhibit significant coupling with the natural environment, forming diverse spatial siting patterns in relation to mountains, water, forests, fields, and the sea, reflecting differentiated adaptation to and ingenious utilization of the natural environment. (4) The adaptive mechanism of the landscape of traditional Chaozhou–Shantou villages can be distilled into a three-tiered structure, natural adaptation as the foundation, social adaptation as the framework, and cultural adaptation as the soul, revealing the spatial planning wisdom of the Chaozhou–Shantou people in complex mountain and coastal environments. This study not only deepens our understanding of the human–land relationship in traditional villages of the Chaoshan region but also provides scientific evidence and theoretical support for the holistic preservation of cultural heritage and regional coordinated development. It holds significant practical value for promoting the protection and sustainable development of rural cultural heritage in the Lingnan coastal region. Full article

The new energy vehicle (NEV) industry is a key sector for achieving dual carbon goals and advancing regional green transformation. Its sustainable development depends on the deep coupling of the industrial chain and the innovation chain. Drawing on data from Shaanxi’s NEV industry covering the period 2014–2023, this study employed kernel density estimation (KDE), the entropy weight method, the coupling coordination degree model, and the optimal parameter geographical detector. Specifically, we examine Shaanxi’s national positioning and spatial pattern within the NEV industry, the spatiotemporal evolution of the coupling coordination degree between its industrial and innovation chains, and the key driving factors along with their interaction mechanisms. The results indicate that Shaanxi is situated within the secondary core growth zone of central and western China. Within the province, the industry exhibits a pronounced spatial pattern characterized by single core concentration in Xi’an, contiguous support across the Guanzhong region, and point-like distribution in northern and southern Shaanxi. The dual-chain coupling coordination degree in Shaanxi’s NEV industry has improved steadily, resulting in a four-tier structure comprising core breakthrough, secondary catch-up, weak foundation, and lagging predicament categories. The dominant driving factors are Industrial Agglomeration Degree, Research and Development (R&D) Funding Input, and Resource Utilization Rate. The interaction between Resource Utilization Rate and Integration Degree exerts the strongest effect. Full article

Single-image dehazing requires wide-range visibility estimation and local structure recovery under spatially varying degradation. Existing large-context models improve global reasoning, but they often incur heavy computation or lose sensitivity to irregular haze boundaries and attenuated details. To address these issues, we propose the Adaptive Deformable Peripheral Convolution Network (ADPCNet), a compact encoder–decoder that organizes dehazing into four coupled operations: conditional adaptive sharing for peripheral large-kernel context modeling, deformable sampling for geometry-aware aggregation, frequency-guided modulation for detail compensation, and dynamic multi-branch fusion for content-adaptive integration. The key idea is to separate broad haze estimation, structure alignment, and detail recovery within an efficient operator stack. Experiments on RESIDE, Dense-Haze, and NH-Haze show that ADPCNet achieves competitive paired-benchmark performance with 7.25 M parameters and 33.62 G FLOPs, reaching 40.89 dB/0.997 on SOTS-Indoor, 37.80 dB/0.996 on SOTS-Outdoor, 18.05 dB/0.679 on Dense-Haze, and 21.66 dB/0.815 on NH-Haze. The ablation and sensitivity results further support the contributions of the proposed modules and the selected kernel configuration. Overall, these results indicate that ADPCNet maintains a favorable quality-efficiency trade-off under the matched paired evaluation protocol. Full article

Microseismic early warning for roof disaster in excavated coal roadways often suffers from low pertinence and a high false positive rate. This study establishes an intelligent early warning process based on unsupervised learning and a voting mechanism. True triaxial compression and drilling tests were conducted to characterize the acoustic emission responses of coal and rock during fracture. Using 720 h of field microseismic data from a high-gas mine in Shanxi, high-weight precursor features were extracted from time–frequency indicators. Kernel principal component analysis (KPCA) was used to optimize the indicator system, and 49 indicators with weights above 0.08 were selected as model inputs. Five unsupervised clustering algorithms were integrated to establish an ensemble decision-making early warning model. The results show that the model eliminates the drawbacks of single algorithms, achieves accurate roof disaster warning, and correctly distinguishes disaster events from non-disaster high-energy events. The false positive rate is zero on the 720 h field dataset, and the reliability of early warning is significantly improved. This study enhances the reliability of mine roof microseismic warning, enriches roof disaster prediction theories, provides a complete intelligent early warning process for mine roof disaster, and offers important references for deep mining dynamic disaster warning research. Full article