Search Results (8,098)

Across four decades of AI deployment, the same six human challenges (trust calibration, reliance behavior, cognitive engagement, skill retention, accountability, and transparency) recur, yet fragmentation across research communities obscures this continuity and limits knowledge transfer. Functionally similar phenomena are repeatedly relabeled (a jangle fallacy): what aviation researchers call “automation complacency,” decision scientists call “algorithm appreciation,” and LLM researchers describe as “over-reliance.” This systematic review synthesizes 152 papers spanning aviation, healthcare, manufacturing/supply chain, and cross-domain contexts across three AI technology generations: decision support systems, autonomous systems, and large language model (LLM) agents. We introduce the Collaboration Convergence Framework (CCF), a 6 × 3 matrix with solution-maturity indicators that maps each challenge across generations. The framework shows that Gen 3 designers can transfer decades of evidence from automation and decision support research (particularly reliance calibration, cognitive forcing, and skill maintenance) rather than rediscovering them. Cross-generational synthesis also isolates three Gen 3 phenomena without direct precedent in earlier generations: epistemia (attributing genuine knowledge to LLMs based on surface fluency), attribution ambiguity in co-creation, and motivational withdrawal. We distill twelve transferable design principles and propose ten research directions, prioritizing skill-retention interventions and accountability frameworks. These findings carry direct sustainability implications aligned with Industry 5.0: protecting workforce capability under increasing automation (SDG 8), reducing duplicated research effort through cross-generational knowledge reuse (SDG 9), and supporting responsible deployment by treating collaboration risks as predictable rather than novel (SDG 12). The CCF provides conceptual infrastructure for cumulative learning across AI generations and industries. Full article

Accurate mapping of winter rapeseed cultivation areas is crucial for food security assessment and agricultural resource management, yet remains a persistent challenge in mountainous regions characterized by complex topography and highly fragmented field parcels. To address these challenges, this study develops a progressive deep learning framework using single growing-season data from the Hanzhong Basin. We conducted a structured comparison of remote sensing indices, machine learning, and deep learning approaches for rapeseed identification in heterogeneous landscapes. First, sensitivity analysis of the Flowering Index for Rapeseed was performed to identify the optimal parameterization, yielding high inter-class separability (ND = 0.959) during peak flowering and a threshold-based overall accuracy (OA) of 94.41%. Second, a multidimensional feature space was constructed by integrating Sentinel-2 spectral bands, image texture metrics, and topographic variables; Random Forest-based feature importance selection subsequently enhanced Support Vector Machine classification performance to an OA of 90.70%. Third, we proposed an innovative three-stage progressive UNet++ architecture: Stage1 focuses on binary rapeseed/non-rapeseed classification to establish spatial priors; Stage2 refines discrimination among spectrally similar vegetation classes (rapeseed and other vegetation); and Stage3 achieves comprehensive seven-class semantic segmentation. A weighted focal loss function combined with a weight inheritance mechanism was employed to mitigate class imbalance and facilitate inter-stage knowledge transfer. The final model attained an OA of 98.65% and a mean intersection over union of 95.29%, while effectively suppressing salt-and-pepper noise artifacts in geometrically fragmented parcels. Our findings demonstrate the substantial advantages of progressive deep learning strategies for crop monitoring in topographically constrained environments. Full article

This article examines gender inequalities in scientific research in the Mediterranean, with a particular focus on STEM disciplines. It draws on qualitative data from the STEP (STEM and Equality, Diversity and Inclusion for Research Enhancement in Portugal) project, a European Commission-funded initiative aimed at embedding equality, diversity, and inclusion (EDI) principles across partner institutions in Portugal, Italy, France, and Spain. Using semi-structured interviews with five scientific leaders and an inductive thematic analysis, the study explores early-stage mechanisms in the institutionalisation of EDI policies and women’s empowerment trajectories from an intersectional perspective. The analysis identifies emergent patterns suggesting: (i) a gradual strengthening of EDI mainstreaming in contexts with initially limited awareness; (ii) the role of transnational collaboration in enhancing visibility, mentoring, and peer learning; and (iii) the potential of time-bounded initiatives to catalyse participant-observed shifts and institutional routines in formation. Rather than measuring longitudinal impact, the article traces how legitimation, routinisation, and network diffusion may enable EDI principles to extend beyond project lifespans and become embedded in governance structures. These mechanism-focused insights offer a transferable framework for future European cooperation initiatives and contribute to ongoing debates on sustainable gender and EDI policy implementation in Mediterranean research contexts. Full article

Accurate identification of weak nodes is a prerequisite for online security assessment, preventive control, and resilience enhancement in modern power systems. However, conventional single-layer graph-learning models mainly emphasize local neighborhood aggregation and are insufficient for characterizing vulnerability propagation from equipment-level disturbance to regional congestion and system-level transfer constraints. This paper proposes a mechanism-aware hierarchical graph-learning framework for weak-node identification in complex interconnected power grids. We emphasize that attention, fusion, and gating operations are standard neural-network mechanisms and are not claimed as new generic deep-learning blocks. The contribution of this paper is the power-system-specific formulation: constructing an electrically meaningful local-supernode hierarchy, defining reproducible mechanism-based node and branch-vulnerability proxies, and interpreting weak-node rankings through node–line–corridor coupling evidence. In the validated implementation, a local graph convolutional encoder and a supernode/global graph convolutional encoder generate 32-dimensional local embeddings and 16-dimensional global embeddings, which are concatenated and decoded by a 48 → 24 → 1 multilayer perceptron to obtain node vulnerability scores. Experiments are conducted on reproducible IEEE benchmark data generated from pandapower standard systems, with representative comparisons on the IEEE 57-bus, 145-bus, and 300-bus systems and a detailed structural interpretation on the IEEE 145-bus case. The present results validate the ability of the implemented local–global hierarchical model to reproduce the proposed mechanism-based vulnerability proxy on representative small- and medium-scale benchmarks. Full article

Cross-condition remaining useful life (RUL) prediction of rolling bearings is affected by distribution shifts between operating conditions, limited labeled target-domain degradation samples, and interference from long stationary healthy stages. Under an offline full-life retrospective analysis protocol, this paper proposes a Degradation-Stage-Aware Transformer-GRU (DSA-TGRU) method. First, a health indicator is constructed from selected multidimensional degradation features by principal component analysis (PCA-HI), and an adaptive threshold moving rate of change (ATMROC) criterion is used to identify the transition from the healthy stage to the degradation stage, defined as the first prognostic time (FPT), i.e., the degradation-start time. Only post-FPT windows are then used to construct RUL labels for model training and evaluation. The prediction model combines a Transformer encoder for long-range sequence dependencies with gated recurrent units for temporal degradation evolution. The model is pretrained on source-domain bearings and then fine-tuned using a small number of labeled target-domain degradation samples available under the offline protocol. Stage-binned sampling and late-stage linear weighting are treated as auxiliary training strategies rather than universally effective modules. Experiments on the XJTU-SY and PHM2012 datasets show that post-FPT degradation modeling and target-domain fine-tuning play major roles in reducing cross-condition errors. The proposed method achieves average normalized MAE values of 0.0492 and 0.0738 and average normalized RMSE values of 0.0626 and 0.0928 on the two datasets, respectively, and generally outperforms several transfer-learning baselines in normalized error metrics. Ablation results further indicate that the benefits of stage-binned sampling and late-stage weighting are dataset- and task-dependent. The current version is not designed for online RUL prediction from incomplete target-bearing trajectories. Full article

Background: Data on assisted reproductive technology (ART) outcomes among Roma women are virtually absent from the literature, despite Roma being the largest and most socioeconomically marginalized ethnic minority in Europe. This study provides the first structured evaluation of IVF/ICSI outcomes among Roma women at a tertiary fertility center. Methods: A retrospective observational cohort study was conducted at the Clinic for Gynecology and Obstetrics, University Clinical Center Niš, Serbia (May 2010–September 2015). Roma (n = 88) and non-Roma women (n = 1197) undergoing IVF/ICSI were compared on baseline clinical, hormonal, and embryological parameters. Primary and secondary outcomes were clinical pregnancy and live birth, respectively. Multivariable logistic regression, propensity score matching (1:4, by age and AMH), first-cycle sensitivity analysis, and a machine learning pipeline (logistic regression, random forest, XGBoost) with SHAP interpretability analysis were applied. Results: Roma women were significantly younger (31.9 ± 4.0 vs. 34.5 ± 4.7 years; p < 0.001) and had a more favorable ovarian reserve profile (AMH 3.78 vs. 2.90 ng/mL; p = 0.004; FSH 6.87 vs. 8.23 IU/L; p < 0.001), yet had a markedly longer duration of infertility (9.3 vs. 6.3 years; p < 0.001). Clinical pregnancy rates (48.9% vs. 41.3%; p = 0.179) and live birth rates (28.4% vs. 30.9%; p = 0.720) were comparable between groups. In multivariable logistic regression and propensity score-matched analyses, Roma ethnicity was not an independent predictor of either outcome. XGBoost SHAP analysis ranked Roma ethnicity last (11th of 11) in feature importance for both clinical pregnancy (mean |SHAP| = 0.033) and live birth (mean |SHAP| = 0.009). The dominant predictors were the number of embryos transferred, AMH, and age. Only 88 Roma women accessed ART over the decade-long study period, indicating profound underutilization of fertility services. Conclusions: No independent association was detected between Roma ethnicity and IVF/ICSI outcomes within the statistical power afforded by the Roma subgroup (n = 88). An exploratory first-cycle live birth signal (adjusted OR = 0.478; 95% CI 0.249–0.920; p = 0.027), not replicated in primary or propensity-matched analyses, is interpreted as hypothesis-generating. The extreme underutilization of ART services among Roma women remains the most clinically salient observation and a priority for targeted public health intervention. Full article

Heap bioleaching is widely used to extract copper from low-grade sulfide ores thanks to its operational simplicity, low cost, and environmental sustainability. However, current control strategies rely primarily on single-factor optimization and often overlook the synergistic interactions of multiple key parameters, such as ore particle size, pore structure, pH, temperature, microbial activity, and oxygen transfer efficiency. As a result, issues such as low recovery rates, extended leaching periods, and high operational costs persist. Moreover, the “gray-box” nature of heap systems impedes real-time monitoring of internal physical, chemical, and biological processes. In addition, empirical multi-parameter optimization is time-consuming and inadequate for capturing complex interdependencies. This review was conducted to systematically examine the key factors influencing heap bioleaching efficiency and critically evaluate recent advances in numerical simulation and intelligent control strategies. As a result, we identified a major research gap: the existing models—including microscale shrinking core models (SCMs), mesoscale pore-network models based on CT reconstruction, and macroscale continuum models—have inherent limitations. SCMs assume idealized spherical particles with uniform mineral distribution while neglecting pore structure evolution and biofilm dynamics. Mesoscale models offer detailed pore characterization but lack robust multi-physics coupling (thermal–hydro–mechanical–chemical–biological, or THMCB). Macroscale models rely on homogenization assumptions that oversimplify spatial heterogeneity and temporal variations in permeability. This analysis covers the relevant literature from 1985 to 2025, with a focus on three methodological scales (micro, meso, and macro) and their integration with machine learning approaches. A notable finding is that hybrid neural network models (e.g., BP and RBF architectures) outperform purely physics-based models in predicting leaching kinetics under varying operational conditions. However, their accuracy depends heavily on high-quality field data—a limitation rarely addressed in prior reviews. By clearly delineating these model-specific limitations and scale-dependent trade-offs, this review makes two unique contributions: a structured framework for selecting and coupling numerical methods according to process requirements and a roadmap for integrating artificial neural networks with multi-physics simulations to achieve real-time intelligent control of heap bioleaching. The findings offer both theoretical guidance and practical references for optimizing the processing of low-grade copper sulfide ores. Full article

Developing translation competence among non-English-major students at Chinese universities remains a pedagogical challenge, especially given the rising demands of cross-cultural communication. This quasi-experimental study examined whether first-year medical students at a Chinese university could improve their translation skills using the constructivist–objectivist theoretical approach (COTA), which combines constructivist learning theories (e.g., active student participation, collaboration, analysis of real-world issues) with objectivist learning methodologies (e.g., sequential skill development, explicit knowledge transfer). In total, 110 students participated in this mixed-methods study. The research methods included (a) pre- and post-tests of students using College English Test Band 4 criteria to evaluate vocabulary, grammar, and accuracy; (b) student perception surveys; (c) semi-structured interviews with instructors; and (d) classroom observations of students, using Gagné’s nine instructional events to ensure faithful implementation of the COTA framework. The COTA-trained students showed statistically significant improvements in translation skills compared to the control group. Additionally, increased student participation and engagement, positive attitudes toward learning, instructors’ ability to implement COTA effectively, and areas for future development were identified in the qualitative findings. These results suggest that integrating constructivist and objectivist teaching philosophies can benefit curriculum designers, language and translation instructors, and policymakers aiming to enhance translation education in Chinese universities and other Asia-Pacific institutions. However, the modest sample size from a single institution limits generalizability, and future studies with larger, more diverse samples are recommended. Full article

County economic growth in multi-tiered fiscal systems depends not only on the volume of transfers but also on whether those transfers pass through intermediary governments. This paper separates administrative delegation from fiscal chain redesign in province-managed county reforms in China. We study 1537 counties from 2000 to 2023 and compare D2, which creates direct province–county fiscal accounts, with D1, which delegates administrative authority but keeps the prefectural intermediary. The empirical design uses panel difference-in-differences estimators, synthetic difference-in-differences, double machine learning robustness checks, and exploratory heterogeneity diagnostics. Based on a placebo-corrected lower bound and a cross-estimator upper bound, D2 is associated with a conservative growth range of 0.35 to 1.0 percentage points per year, while the D1 estimate is imprecise. D2 is also associated with higher contemporaneous per capita fiscal expenditure, but the one-year lagged mediator check does not support a fully identified expenditure mechanism. Heterogeneity patterns are consistent with stronger effects in transfer-dependent counties, but they remain exploratory. The outcome is county economic growth, not a composite sustainability index. The results support a focused governance claim. More reliable transfer delivery is consistent with improved local growth capacity, while fiscal, social, and environmental sustainability remain outside the measured outcome space. Full article

The emergence of citizen energy communities under the European Clean Energy Package is creating new opportunities for neighboring households to collectively reduce electricity costs through local energy sharing. This paper presents a distributed multi-agent energy management system for a two-household residential energy community in which each household is equipped with photovoltaic generation and a battery energy storage system operating under realistic hourly-varying electricity prices. Each household is managed by an independent Deep Q-Learning agent that learns a cost-optimal charging and discharging policy using only local observations. In parallel, a coordination agent, implemented on the SPADE platform with XMPP-based messaging, oversees real-time peer-to-peer energy transfers between households, enabling energy exchange whenever one household has surplus generation and another faces a deficit. The two households are deliberately configured with complementary profiles: one has higher PV generation capacity while the other has higher energy consumption. This setup creates natural opportunities for local energy sharing between them. Performance is assessed through a three-level evaluation framework: (i) individual household economics (cost reduction, battery management, grid exchanges), (ii) coordination efficiency (transfer frequency, direction, and volume), and (iii) aggregate community performance, which isolates the added value of peer-to-peer sharing beyond what each household achieves through individual BESS optimization. Numerical experiments using GEFCom2014 solar generation data, synthetic residential load profiles calibrated following documented consumption patterns, and day-ahead price signals representative of the Spanish electricity market demonstrate that both Deep Q-Learning agents independently learn effective charge/discharge strategies aligned with price signals and PV availability. They also show that the coordination layer further reduces community grid dependence by routing surplus energy locally rather than exchanging it with the main grid at less favorable rates. The results confirm that a well-engineered integration of decentralized reinforcement learning with a lightweight coordination protocol can deliver measurable economic benefits in realistic residential energy communities without requiring centralized training, shared data, or complex multi-agent reinforcement learning architectures. Full article

Motor imagery (MI)-based brain–computer interfaces (BCIs) have led to great interest as a result of their potential use in neurorehabilitation, assistive robotics, and human–computer interaction. However, decoding electroencephalographic (EEG) signals with high accuracy continues to be a difficult task due to the weak signal-to-noise ratio, differences among subjects, and the complicated temporal–spectral–spatial neural dynamics. Deep learning methods recently developed, such as convolutional neural networks, recurrent architectures, graph neural networks, and adversarial transfer learning, have enhanced MI decoding performance, yet many models are still concentrating on a single representation domain or they need costly adaptation phases in terms of computation. To tackle these shortcomings, we present NeuroCrossNet, a unified tri-modal deep learning model that is able to learn the temporal, spectral, and spatial EEG features jointly for robust and calibration-free MI decoding. The suggested network combines a Temporal HyperMixer Block for capturing long-range temporal dependencies, a wavelet transformer for learning localized time–frequency representation, and a Graph Attention Network for EEG topology-aware spatial reasoning. Additionally, a Dynamic Residual Attention Gate (DRAG) has been developed to adaptively merge heterogeneous feature streams, and a compact subject-aware normalization (SAN) method enhances cross-subject generalization without the use of labeled target-domain calibration data. Our proposed model was tested following the rigorous leave-one-subject-out (LOSO) approach on BCI Competition IV-2a and High-Gamma datasets. NeuroCrossNet reached a classification accuracy of 91.30%, surpassing several strong benchmark methods, including CNN-LSTM, EEGNet, DeepConvNet, spectral CNN, and graph-based EEG decoding frameworks. Furthermore, a large number of ablation studies reveal that the integration of temporally, spectrally, and spatially complementary representations considerably boosts robustness and inter-subject consistency. Full article

Aiming at the problems of extremely scarce measured samples and significant domain shift between simulated and measured data in automatic target recognition (ATR) of air targets for spaceborne radar, this paper proposes an inverse synthetic aperture radar (ISAR) image recognition method for air targets combining physics-driven data augmentation guided by detection prior information with domain adversarial transfer learning. First, the mapping relationship between scattering point projection and ISAR images is established by using the target 3D point cloud and radar observation geometric priors, and a 2D sinc kernel function is introduced for energy distribution rendering. Then, under the unsupervised transfer learning paradigm, aiming at the distribution inconsistency between augmented data (source domain) and unlabeled simulated data (target domain), this paper designs a cross-domain recognition task experiment including six types of typical aircraft targets, and compares the cross-domain recognition performance of three transfer learning methods (model fine-tuning, deep domain confusion (DDC) and domain-adversarial neural networks (DANN)) on the target domain. Meanwhile, t-distributed stochastic neighbor embedding (t-SNE) visualization is used to analyze the feature distribution alignment ability of the models. Simulation experiments show that the DANN model with a dynamic inversion coefficient introduced in the gradient reversal layer (GRL) achieves a recognition accuracy of 99.5% on the unlabeled target domain, which is significantly superior to the model fine-tuning and DDC methods. Moreover, it makes the feature distributions of source and target domain samples highly overlapping, and maintains a strong inter-class discriminability while eliminating the domain shift. The proposed scheme provides a physically interpretable and robust technical path for few-shot radar target image recognition. Full article

The purpose of this review is to present an overview of artificial intelligence methods for classifying paintings into the artistic movement to which they belong. To achieve this goal, a literature review of research articles from the 2014–2024 period was carried out. The search for scientific articles was carried out in the Scopus database. The initial search yielded 492 publications and after successive stages of screening and full-text evaluation, 39 articles were finally selected for detailed analysis. The review presents (a) the datasets used in the works, (b) the range of artistic movements examined and (c) the computational methods from machine learning to deep neural networks and transfer learning. Methodological issues are highlighted, such as class imbalance of the samples, dataset bias and the limitations of commonly used evaluation metrics. The general finding is that a variety of methodologies were applied, with an increasing use of deep learning and transfer learning models, which in many cases are reported as effective within specific datasets and experimental protocols. Finally, the review offers a taxonomy of methodologies and maps trends and research gaps in research on painting style classification over the last decade, while at the same time making suggestions for future research. Full article

Background/Objectives: Accurate land cover mapping from multispectral Sentinel-2 imagery is fundamental for environmental monitoring, efficient natural resource management, and spatial planning. While deep learning has become the dominant approach for semantic segmentation, the combined impact of spectral input selection and network architecture on cross-regional robustness remains insufficiently explored. This study systematically investigates multispectral land cover segmentation in Serbia and evaluates its transferability to Western Balkan regions using a structured experimental framework. Methods: A comprehensive band-combination ablation analysis (3–10 spectral bands and index-only inputs) was first conducted using Attention U-Net, followed by a comparative evaluation of representative convolutional and transformer-based architectures, including ResNet-UNet-50, ConvNeXt-UNet, DeepLabV3+ (ResNet-50), and DINOv2-S/14. Model performance is evaluated on an internal Serbian test split (Test SR), an external Serbian dataset (Ext SR), and a cross-regional Balkan dataset (Ext WB). Results: The results demonstrate that compact multispectral configurations (6–9 bands) provide the most stable performance, achieving mIoU values of approximately 0.72–0.74 under in-domain evaluation and remaining robust under external testing. The inclusion of near-infrared and shortwave infrared bands proved critical for effective land cover discrimination, whereas increasing spectral dimensionality beyond this range did not yield systematic improvements in external robustness. Notably, the magnitude of performance degradation under pronounced geographic domain shift exceeds the performance differences observed between architectures under in-domain conditions, indicating that distribution shift exerts a stronger influence on segmentation accuracy than model choice alone. Class-wise analysis revealed agricultural areas as the most domain-sensitive category, while Shapley-based explainability analysis provides additional insight into class-specific spectral dependencies and their role in generalization behavior. Conclusions: Although transformer-based models demonstrated competitive robustness, attention-enhanced convolutional architectures achieved comparable stability across evaluation scenarios. Overall, the findings emphasize the importance of balanced spectral design, class-aware robustness analysis, and explicit out-of-domain evaluation for developing transferable land cover segmentation models in remote sensing applications. Full article

Transfer learning offers a practical way to improve neural machine translation in low-resource settings, but its effectiveness depends on both the choice of transfer language and the amount of target-language data available for adaptation. In this study, we examine these factors specifically for Polish–English translation using mBART. We evaluate Czech, Russian, and German as parent languages and extend the analysis with a combined Slavic parent model trained on Czech and Russian. The models are compared across 0-shot, 10-shot, 100-shot, 1k-shot, and 10k-shot settings. Within this Polish–English mBART setting, Czech provides the strongest zero-shot performance, while Russian and German improve substantially as Polish fine-tuning data increases and achieve the strongest results at higher shot levels. The paper therefore analyzes selected transfer-language configurations rather than a formally measured similarity variable. The results suggest that, in this setup, transfer-language choice matters most when no Polish supervision is available, whereas larger amounts of Polish data can compensate for weaker initial transfer alignment. Full article