Search Results (2,843)

Multi-view clustering has emerged as a significant research direction in the information age, as multiple feature representations become increasingly available. However, traditional multi-view clustering methods are often unsupervised and fail to exploit available label information. In practice, fully labeled data are scarce, while partially labeled data are more common and can significantly improve clustering performance. Moreover, real-world data are frequently corrupted by noise or outliers. To address these challenges, this paper proposes a Correntropy-guided Matrix Factorization and Tensor Graph Learning (CMTGL) framework for robust semi-supervised multi-view clustering. Specifically, CMTGL incorporates partial label information into a shared low-dimensional representation through constrained low-rank matrix factorization and employs the maximum correntropy criterion (MCC) to reduce the influence of noisy samples and outliers. In addition, view-specific graphs are stacked into a third-order tensor and regularized by the tensor Schatten p-norm to exploit high-order correlations across multiple views. A consensus graph is jointly learned to capture the common structural information shared among different views. The resulting optimization problem is efficiently solved by a block coordinate descent algorithm with an extrapolation accelerated block coordinate update (BCU) scheme. Extensive experiments on six public benchmark datasets demonstrate that the proposed method achieves superior clustering performance compared to state-of-the-art approaches. Full article

Perceived safety determines whether urban greenways are used in everyday life, yet it is rarely measurable at the boundary scale where design decisions are made. Existing street-view models split into black-box networks whose predictions cannot be traced to design elements and pixel-ratio regressions whose interpretability rests on weak, unstructured representations, while greenspace studies lean on GIS proximity variables that confound design with context. We present the CPTED-Guided Perception Network (CGPN), which fuses a visual branch with a masked, learnable projection of segmentation ratios onto five CPTED dimensions. Because the mask confines learning to a theory-defined support, the prior regularizes the representation while every coordinate of the model remains tied to a named CPTED dimension, whose directional effect on the prediction we verify by perturbation. On 110,633 street-view images, CGPN is statistically equivalent to the strongest black-box baseline in pairwise ranking accuracy (0.649 vs. 0.652; equivalence test within a 1.5-point margin, $p=0.006$, attains the best $R^2$ (0.192), and improves on its unconstrained variant in goodness of fit across three seeds ($\mathrm{\Delta }{R}^2=+0.031$, $p=0.042$). Applied to 218 greenway-adjacent residential boundaries in Boston and New York, it uncovers a threshold-like negative association for barrier-dominated access control and an inverted-U distance profile whose weakest segment lies within 100 m of the greenway edge ($p=0.007$). Full article

Multimodal Sequential recommendation alleviates the semantic insufficiency and data sparsity of item-ID-based models by incorporating side information such as text and images. However, multimodal systems face the dual challenges of feature-space heterogeneity and modality-specific noise, in addition to the dynamic evolution of user interests over time. Existing methods still struggle to jointly handle cross-modal alignment and time-aware preference modeling. To address these challenges, we propose a multimodal sequential recommendation framework with cross-modal alignment and temporal gating, which leverages item ID, text, and image modalities to capture users’ dynamic interests. The proposed model contains three core components. First, a cross-modal alignment mixture-of-experts module preserves modality-specific features with dedicated experts and captures shared semantics with common experts, thereby mitigating the semantic mismatch inherent in direct fusion. Second, a hierarchical time-aware mixture-of-experts module uses short-term intervals, long-term spans, and periodic time encodings for expert routing, and applies a time-aware modality gate to adaptively adjust the importance of ID, text, and image modalities under different temporal contexts. Third, a sequential interest contrastive learning objective enhances the discriminability of ID-based sequential interest representations by leveraging dynamic temperature scaling, multi-scale positive samples, hard negative mining, and diversity regularization. Experiments on games, beauty, and toys demonstrate that the proposed method consistently outperforms representative sequential and multimodal recommendation baselines on Normalized Discounted Cumulative Gain (NDCG)@5, NDCG@10, Mean Reciprocal Rank (MRR)@5, and MRR@10. Furthermore, ablation results validate the effectiveness of each proposed component. Full article

Reliable attribution of artificial intelligence (AI)-generated text to a specific large language model (LLM) matters increasingly as LLMs proliferate, yet where quantum machine learning actually stands on this task has, to our knowledge, never been measured systematically. We benchmark the quantum support vector machine (QSVM) for binary attribution between Gemma 3 and Qwen 2.5 on a 5800-sample corpus from paired prompts: 83 configurations sweeping qubit count, regularization, training-set size, feature-map family, and circuit depth under exact, noiseless classical statevector simulation. QSVM validation accuracy plateaus at approximately 88%, whereas a classical support vector machine with a radial basis function kernel reaches approximately 97.8% on the identical fourteen-dimensional inputs: the ceiling belongs to the quantum (fidelity) kernel, not to the input representation. We measure the mechanism: off-diagonal quantum kernel values shrink exponentially with qubit count, the signature of exponential kernel concentration. The same classical model recovers the stylometric attribution fingerprint, showing it belongs to the shared feature pipeline rather than to the quantum kernel. All large-scale headline results generalize to an independent 1000-text test set produced after every design decision was frozen. The study provides a cautionary, reproducible benchmark for quantum kernel natural language processing and outlines an open-set extension as future work. Full article

Reliable long-term city-scale nitrogen dioxide (NO₂) records are essential for evaluating urban air quality change, but the short observation period of TROPOMI limits long-term applications. This study developed a TROPOMI-referenced machine learning framework to reconstruct annual tropospheric NO₂ column density for 436 cities in China, Japan, South Korea, North Korea, and Mongolia from 2000 to 2022 using 18 annual predictors comprising five natural environmental variables, six meteorological variables, and seven sectoral anthropogenic NO_x emission variables. To reduce spatial leakage, the 436 city polygons were assigned to a regular 5° × 5° grid using the largest equal-area polygon intersection fraction. The 61 occupied, non-overlapping blocks were allocated deterministically to five folds, with all 2019–2022 observations from each city retained in its assigned block. Model performance was calculated from the concatenated predictions for the five held-out block sets. Among nine models, random forest (RF) achieved the best independent test performance (R² = 0.885; RMSE = 1.968 × 10⁻⁵ mol m⁻²). During 2005–2022, annual city-level RF–OMI correlations ranged from 0.850 to 0.945, and their normalized regional annual series were strongly correlated (r = 0.877). Against ground observations, RF better represented intercity differences in China (mean annual r = 0.791 versus 0.735 for OMI), whereas OMI performed better at the Japanese city scale (0.857 versus 0.810 for RF); nevertheless, RF closely reproduced the Japanese national annual trend (r = 0.989). The East Asian mean increased significantly during 2000–2011 (Theil–Sen slope = +0.095 × 10⁻⁵ mol m⁻² yr⁻¹), declined significantly during 2011–2018 (−0.140 × 10⁻⁵ mol m⁻² yr⁻¹), and remained nonsignificantly negative during 2018–2022 (−0.060 × 10⁻⁵ mol m⁻² yr⁻¹). China peaked in 2011, Japan and North Korea showed significant long-term decreases, South Korea showed a significant overall decline, and Mongolia had no significant full-period trend. SHAP analysis showed that industrial combustion emissions, surface pressure, and road emissions had the highest global mean absolute SHAP values, with nonlinear and direction-dependent associations with RF predictions. The resulting dataset supports regional and national long-term NO₂ assessment, while country-specific and city-scale uncertainties should be considered in local applications. Full article

Background: Autoimmune hepatitis (AIH) is a chronic immune-mediated liver disease that may progress to cirrhosis. This study aimed to develop and externally validate interpretable machine-learning models for the classification of prevalent cirrhosis in patients with AIH. Methods: The development cohort included 55 patients with biopsy-confirmed AIH. Cirrhosis was defined histologically as F4, whereas F0–F3 was classified as non-cirrhosis. Logistic Regression with L2 regularization, Random Forest, and XGBoost were evaluated. The original stratified 70/30 hold-out analysis was retained, and repeated stratified five-fold cross-validation with 20 repeats was additionally performed to assess internal stability. Primary external validation was performed in an independent histology-matched cohort of 42 patients. Models were applied without refitting, recalibration, or threshold optimization. Discrimination, probabilistic accuracy, calibration, and threshold-dependent classification metrics were evaluated. Results: In the original held-out test set, AUROC was 0.900 for Logistic Regression with L2 regularization, 0.830 for Random Forest, and 0.890 for XGBoost. In repeated cross-validation, mean AUROC was 0.878 ± 0.021, 0.914 ± 0.015, and 0.896 ± 0.015, respectively. In the primary external validation cohort, Random Forest showed the highest numerical discrimination (AUROC 0.810; 95% CI, 0.653–0.933), followed by XGBoost (0.728; 95% CI, 0.566–0.878) and Logistic Regression with L2 regularization (0.716; 95% CI, 0.545–0.875). Random Forest also had the lowest Brier score (0.188). However, confidence intervals were wide and overlapping. Elastography stage alone achieved an AUROC of 0.745, and the numerical improvement of the full Random Forest model was not statistically clear. Conclusions: Machine-learning models integrating routinely available clinical, biochemical, immunological, and elastography-related variables showed preliminary external transportability for the classification of prevalent cirrhosis in AIH. However, the small development and validation cohorts, uncertainty in calibration, and lack of a clearly demonstrated incremental advantage over elastography alone indicate that larger prospective multicenter studies are required before clinical implementation. Full article

The power demand of High-Performance Computing (HPC) infrastructures exhibits both stable weekly regularities and rapid workload-driven fluctuations, which are difficult to capture reliably with a single modeling paradigm. Achieving more sustainable HPC operation requires accurate forecasts at multiple horizons: short-term predictions support operational control (e.g., proactive power capping and energy-aware scheduling), whereas long-term forecasts are essential for planning activities (e.g., capacity provisioning and energy procurement). Together, these capabilities reduce operational cost and risk while enabling more efficient and sustainable datacenter management. This paper investigates multi-horizon forecasting of aggregated active power consumption in an operational HPC datacenter utilizing a four-month dataset (five-minute time intervals) from the University of Turin. We propose a hybrid residual-learning framework that integrates a long-term structural forecaster with a short-term residual corrector utilizing a Temporal Convolutional Network (TCN) to address the simultaneous presence of weekly regularities and short-term workload-induced fluctuations. Assessment utilizing a rolling-origin protocol covers a timeframe of 15 min to 6 h and extends 1 to 3 weeks into the future. Performance of the proposed approach has been compared against the SARIMA baseline. Full article

Aluminum alloys used in aircraft structures are susceptible to corrosion fatigue cracking under combined aggressive environments and cyclic loading, threatening structural integrity. Pure data-driven models often fail under distribution shifts, while single-physics-informed neural networks (PINNs) lack flexibility in complex conditions. This paper proposes a dual-physics-informed neural network (DPINN) that integrates Walker and Forman crack growth models into a deep residual network. The model adaptively fuses both physical formulas via a trainable weight α and predicts material constants. A hybrid loss function with α regularization ensures physically consistent predictions. Using comprehensive corrosion fatigue data covering eight aluminum alloys, we evaluate the model on an internal test set and, more importantly, on an independent external test set simulating real-world distribution shifts. We further investigate an incremental learning scenario where the model is sequentially fine-tuned with increasing fractions of the external set. Results demonstrate that DPINN rapidly rectifies initial distribution mismatch, crossing the engineering reliability threshold (R² > 0.90) at an early incremental stage, and achieves superior performance after fine-tuning, significantly outperforming both a single Walker-PINN and gradient boosting regressors. SHAP feature importance analysis identifies ΔK and stress ratio as dominant drivers, confirming mechanistic consistency. The proposed architecture offers a data-efficient and interpretable tool for corrosion fatigue crack growth prediction in aluminum alloy structures. Full article

Maritime object detection is essential for port-based surveillance, ship tracking, and maritime security. However, practical implementation encounters two primary challenges: significant class imbalance (IMO identification numbers represent only 1.2% of labelled objects) and spatial inconsistency (predicted IMO numbers often appear beyond ship boundaries). Standard detectors and conventional class-balancing strategies fail to adequately address these difficulties, resulting in a persistent mismatch between research validation and practical performance. We present a hybrid autoencoder–YOLO framework with variable spatial regularisation. To the best of our understanding, this is the first methodology to simultaneously address class imbalance and spatial inconsistency in maritime IMO detection via reconstruction-guided learning and differentiable spatial regularisation. The model uses a YOLOv8 encoder shared by both a detection head (designed for ships and IMO numbers), and an auxiliary reconstruction decoder (a SkipDecoder with U-Net-style skip connections). A unique spatial limitation loss provides the physical restriction that each IMO number must be within a ship’s bounding box, using only anticipated boxes. Training occurs in two phases: autoencoder pretraining on the marine dataset, followed by phased joint optimisation with a curriculum schedule for the spatial weighting. In a dataset of 297 annotated images (utilising five-fold cross-validation with a 47-image preserved test set), our comprehensive model achieved 50.1% IMO AP50-95, 97.8% IMO precision, and 94.6% IMO F1-score on the test set, beating the baseline YOLOv8s by +7.8 percentage points, +6.5 percentage points, and +5.2 percentage points, respectively. Ablation studies indicate that reconstruction instruction improves IMO AP50-95 by +4.5 percentage points, while spatial regularisation adds +3.3 percentage points. Although ship detection results in a small compromise (ship AP50-95 decreases from 77.0% to 55.6%), this appears to be practically acceptable given the essential role of IMO numbers as unique ship IDs. Significantly, inference speed improves by 20% (8.0 ms per image on an NVIDIA A100 GPU) relative to YOLOv8s (10.0 ms). Comparisons with leading detectors (RetinaNet, Faster R-CNN, DETR, EfficientDet), all initialised with standard COCO-pretrained backbones and fine-tuned on our dataset, reveal that none achieve an IMO AP50-95 exceeding 42.3%, highlighting the task’s challenge and the accuracy of our design. The proposed framework shows potential to address the gap in practical implementation through reconstruction-based feature learning accompanied by a specified geometric baseline. It is precise, accurate, and fast, aligns with specific physics, and shows promise for real-time maritime surveillance applications, though we acknowledge the need for additional thorough verification across many operational environments. Full article

Reliable abnormal behavior recognition from surveillance videos is hindered by the high cost of clip-level annotation, the scarcity of abnormal samples, and the context-dependent nature of behavioral semantics. Although vision–language models offer strong semantic transferability, their application under limited supervision remains susceptible to noisy pseudo-labels and confirmation bias. We propose confidence-aware semi-supervised vision–language contrastive learning (CA-VLC), which jointly exploits limited labeled videos and abundant unlabeled videos. Building on an existing CLIP-initialized temporal backbone, CA-VLC combines behavior-only and context-enriched text prototypes through confidence- and agreement-guided semantic fusion. For unlabeled videos, the model generates predictions from weakly augmented views and selects reliable pseudo-labels using entropy-based confidence estimation and class-adaptive thresholds. Detached weak-view targets then supervise strongly augmented views through confidence-weighted self-training without requiring an additional teacher network. Furthermore, cross-view consistency regularization and confidence-aware contextual alignment suppress unreliable semantic cues and improve robustness to contextual noise. Experiments on CABR50 demonstrate consistent improvements across multiple labeled-data ratios, while evaluations on CABRZ6 and UCF-101 assess prompt-based transfer to predefined target label sets without target-domain fine-tuning. With 10% labeled videos, CA-VLC achieves 84.06% Top-1 accuracy and 83.51% Macro-F1, retaining 95.47% of its fully supervised Top-1 accuracy of 88.05%, thereby demonstrating its effectiveness for label-efficient abnormal behavior recognition. Full article

Person–Job Fit (PJF) serves as a core task of intelligent recruitment recommendation. However, existing graph-based PJF models rely on a fixed, single-path aggregation scheme, thereby failing to simultaneously capture global interaction statistics and local competency-matching signals from candidate–job bipartite graphs. To address this limitation, this work proposes Bi-path Graph Convolutional Neural Network with Gate Fusion (BiGCNG), a dual-path graph convolutional network with global learnable gate fusion, composed of three coordinated modules. First, the Shared Text Embedding Pre-processing Module (STEPM) generates unified node embeddings by fusing structured attributes and BERT contextual text features. Second, the Bi-path Graph Convolution Module (BiGCM) extracts multi-granularity graph representations via separate sum and max aggregation paths. Third, the lightweight Gate Fusion Module (GFM) balances two feature streams via a learnable global scalar gate. The model is optimized with regularized Bayesian Personalized Ranking (BPR) loss on highly sparse recruitment data (99.97% sparsity). BiGCNG is evaluated on the Zhilian dataset, a real-world Chinese recruitment dataset, and outperforms five mainstream baselines notably, increasing MRR@5 by 7.67% and NDCG@5 by 5.48% on the Candidate subset, 2.30% and 0.61% on the Job subset against the best baseline, respectively. Several visualizations and hyperparameter analysis jointly validate the effectiveness and robustness of dual-path propagation and gate fusion. This work provides an effective multi-granularity graph learning paradigm for intelligent talent recruitment matching. Full article

Steganography using deep learning can preserve pixel-level image quality while still changing object, attribute, or relational information in captions generated by vision-language models (VLMs). This caption drift creates a detection channel that is not measured by global image-embedding similarity alone. This paper presents StegoGuard, a framework that embeds secret payloads while enforcing caption-level semantic consistency. Concept drift is defined as a measurable divergence in object-level, attribute-level, or relational semantics between cover and stego captions and is quantified using CLIP text-embedding cosine distance. The main technical contribution is a cross-modal semantic drift regularization term based on BLIP-2 captions generated for cover and stego images. The framework combines this objective with CLIP-based saliency-guided region selection and a lightweight Vision Transformer encoder-decoder. Saliency-map quality is evaluated against ground-truth segmentation masks, and a deterministic bit-to-patch mapping protocol is provided for reproducibility. Experiments use COCO2017, DIV2K, and BOSSBase. Within the controlled six-baseline protocol, StegoGuard achieved a PSNR of 38.9 dB, an SSIM of 0.976 at 256 bits, detector AUC values of 0.521–0.562, and caption similarity of 0.962 as measured by CLIP text cosine similarity (model-relative, not human-verified). The ablation results show that the drift term reduces the measured concept-shift rates while preserving the reported bit-recovery and image-quality levels. Full article

Time-series representation learning decomposes signals into interpretable factors such as trend and seasonality, and disentangled representation learning assigns these factors to distinct latent dimensions, enhancing interpretability and forecasting accuracy. However, existing methods achieve only statistical disentanglement: an intervention on one generative factor can still influence the latent assigned to the other, since unobserved common causes induce dependence between them. In time-series data, this entanglement is amplified by time-varying confounders, distorting causal effect estimates and degrading counterfactual prediction accuracy. We therefore propose CDTS²—Causal Downstreamer with Causally Disentangled Trend and Seasonality Time-Series Representations. CDTS² combines two complementary mechanisms: (i) dedicated subnetworks separate trend and seasonality, regularized by the Hilbert–Schmidt Independence Criterion (HSIC) to suppress residual dependence between the two latents; and (ii) a causal discovery objective infers a summary causal graph over the input variables and regularizes the shared representation underlying both latents to incorporate causal dependency structure during training. To assess causal disentanglement, we evaluate CDTS² on counterfactual prediction in medical and energy domains. Under severe confounding, CDTS² reduces RMSE by 11.6% and 17.1% over the second-best baselines on the tumor-growth and CityLearn datasets, while preserving the best factual-forecast accuracy. CDTS² also attains the highest Interventional Robustness Score (IRS) and Disentanglement–Completeness–Informativeness (DCI) metrics, indicating that it learns causally disentangled representations that remain robust under interventions. Full article

This study presents a leakage-proof hybrid machine learning framework for early flood risk forecasting using multimodal spatiotemporal tabular data. An event-based dataset covering Kazakhstan from 2001 to 2021 was created by integrating topographic characteristics, hydrometeorological variables, long-term surface water dynamics, and remotely sensed spectral indices. To ensure realistic assessment, only pre-event observations were used, and event-based temporal data separation was employed to prevent leakage between the training and test subsets. The proposed Remote Sensing Adaptive Linear Opinion Pool Machine Learning (RS-ALOP-ML) framework combines multiple logistic regression experts with different regularization strengths through an Adaptive Linear Opinion Pool (ALOP) probabilistic fusion strategy, thereby preserving interpretability while improving forecasting robustness. The proposed framework was evaluated for four independent forecast horizons (T + 1, T + 7, T + 14, and T + 30 days) and compared with traditional machine learning algorithms and state-of-the-art tabular deep learning models, including Random Forest, ExtraTrees, XGBoost, LightGBM, CatBoost, MLP, FT-Transformer, TabNet, and Process Wide&Deep. Experimental results demonstrate that the proposed hybrid approach consistently achieves competitive or superior forecasting performance while maintaining computational efficiency and transparent probabilistic outputs. The study highlights that carefully designed hybrid machine learning architectures, combined with leakage-safe evaluation protocols, provide a robust foundation for multimodal environmental forecasting and decision support applications. Full article

Ship detection in Synthetic Aperture Radar (SAR) imagery is essential for maritime surveillance and situational awareness. Despite the advances of deep learning for SAR ship detection, small-target detection is still hindered by severe feature degradation from repeated down-sampling and insufficiently discriminative representations under weak scattering and complex background clutter. To alleviate this dilemma, a Spatial Prior-guided feature enhancement and Multi-branch Complementary learning framework is proposed, termed SPMC, for small ship detection in SAR images. Specifically, a Spatial Prior-Guided Feature Enhancement (SPFE) module is designed to derive spatial attention priors for multi-scale features from ground-truth annotations, thereby emphasizing target-related responses and strengthening small-ship representations. Second, a Multi-branch Complementary Classification (MCC) module is developed, which introduces multiple auxiliary classification heads to learn complementary discriminative information from different classification perspectives. Furthermore, a dual-weighted complementary regularization strategy is proposed to encourages different classifiers to focus on hard samples, thereby improving the discriminative capability for small ships. Extensive experiments on the HRSID and LS-SSDD benchmarks validate the effectiveness of the proposed framework. For extremely small ships with very limited image coverage, SPMC improves the baseline YOLOv11 detector by 3.25%/0.99% in AP₅₀/AP_0.5:0.95 on LS-SSDD and by 1.28%/1.21% on HRSID, demonstrating its effectiveness in challenging SAR small ship detection. Full article