Search Results (146)

Stance detection has emerged as an essential tool in natural language processing for understanding how individuals express agreement, disagreement, or neutrality toward specific targets in social and online discourse. It plays a crucial role in bilingual and multilingual environments, including English-Arabic social media ecosystems, where differences in language structure, discourse style, and data availability pose significant challenges for reliable stance modelling. Existing approaches often struggle with target awareness, cross-lingual generalization, robustness to noisy user-generated text, and the interpretability of model decisions. This study aims to build a reliable, explainable target-aware bilingual stance-detection framework that generalizes across heterogeneous stance formats and languages without retraining on a dataset specific to the target language. Thus, a unified dual-encoder architecture based on mDeBERTa-v3 is proposed. Cross-language contrastive learning offers an auxiliary training objective to align English and Arabic stance representations in a common semantic space. Robustness-oriented regularization is used to mitigate the effects of informal language, vocabulary variation, and adversarial noise. To promote transparency and trustworthiness, the framework incorporates token-level rationale extraction, enables fine-grained interpretability, and supports analysis of hallucination. The proposed model is tested on a combined bilingual test set and two structurally distinct zero-shot benchmarks: MT-CSD and AraStance. Experimental results show consistent performance, with accuracies of 85.0% and 86.8% and F1-scores of 84.7% and 86.8% on the zero-shot benchmarks, confirming stable performance and realistic generalization. Ultimately, these findings reveal that effective bilingual stance detection can be achieved via explicit target conditioning, cross-lingual alignment, and explainability-driven design. Full article

Common-mode voltage (CMV) is a critical concern in motor drive applications employing multilevel inverters, as it can lead to significant issues such as high-frequency noise, electromagnetic interference, and motor bearing degradation. These effects can compromise the reliability, reduce the operational lifespan of electric machines, and introduce safety hazards. In this study, an enhanced Direct Torque Control (DTC) strategy incorporating Space Vector Modulation (SVM) is proposed to specifically address CMV-related challenges in induction motors (IM) driven by a three-level Neutral-Point-Clamped (NPC) inverter. The proposed DTC scheme utilizes a specialized modulation technique that effectively mitigates CMV while also minimizing current harmonic content, and torque and flux ripples with a constant switching frequency. The developed SVM algorithm simplifies the three-level space vector representation into six equivalent two-level diagrams, enabling more efficient control. The zero-voltage vector is synthesized virtually by combining two active vectors within a two-level hexagonal structure. The effectiveness of the proposed DTC approach is validated through both simulation and Hardware-In-the-Loop (HIL) testing. Compared to the conventional DTC method, the proposed solution demonstrates superior performance in CMV minimization and leakage current reduction. Notably, it limits the CMV amplitude to V_dc/6, a significant improvement over the V_dc/2 typically observed with the standard DTC approach. Full article

Multi-view clustering, which improves clustering performance by using the complementary and consistent information from multiple diverse feature sets, has been attracting increasing research attention owing to its broad applicability in real world scenarios. Conventional approaches typically leverage this complementarity by projecting different views into a common embedding space using view-specific or shared non-linear neural networks. This unified embedding is then fed into standard single-view clustering algorithms to obtain the final clustering results. However, a single common embedding may be insufficient to capture the distinct or even contradictory characteristics of multi-view data, due to the divergent representational capacities of different views. To address this issue, we propose a mixture of experts (MoE) based embedding learning method that adaptively models inter-view relationships. This architecture employs a typical MoE module as a projection layer across all views, which uses shared expert and several groups of experts for consistency and complementarity mining. Furthermore, a Kullback-Leibler divergence based objective with over clustering is designed for clustering-oriented embedding learning. Extensive experiments on six benchmark datasets confirm that our method achieves superior performance compared to a number of state-of-the-art approaches. Full article

Glioblastoma (GBM) is the most common and aggressive primary brain tumor in adults, with a median overall survival of fewer than 15 months despite standard-of-care treatment. Accurate preoperative prognostication is essential for personalized treatment planning; however, existing approaches rely primarily on magnetic resonance imaging (MRI) and often overlook the rich histopathological information contained in postoperative whole-slide images (WSIs). The inherent spatiotemporal gap between preoperative MRI and postoperative WSIs substantially hinders effective multimodal integration. To address this limitation, we propose a contrastive-learning-based Imaging–Pathology Synergistic Alignment (CL-IPSA) framework that aligns MRI and WSI data within a shared embedding space, thereby establishing robust cross-modal semantic correspondences. We further construct a cross-modal mapping library that enables patients with MRI-only data to obtain proxy pathological representations via nearest-neighbor retrieval for joint survival modeling. Experiments across multiple datasets demonstrate that incorporating proxy WSI features consistently enhances prediction performance: various convolutional neural networks (CNNs) achieve an average AUC improvement of 0.08–0.10 on the validation cohort and two independent test sets, with SEResNet34 yielding the best performance (AUC = 0.836). Our approach enables non-invasive, preoperative integration of radiological and pathological semantics, substantially improving GBM survival prediction without requiring any additional invasive procedures. Full article

Rapid and accurate acquisition of landslide inventories is essential for effective disaster relief. Deep learning-based pixel-wise semantic segmentation of remote sensing imagery has greatly advanced in landslide mapping. However, the heavy dependance on extensive annotated labels and sensitivity to domain shifts severely constrain the model performance in unseen domains, leading to poor generalization. To address these limitations, we propose LandsDANet, an innovative unsupervised domain adaptation framework for cross-domain landslide identification. Firstly, adversarial learning is employed to reduce the data distribution discrepancies between the source and target domains, thereby achieving output space alignment. The improved SegFormer serves as the segmentation network, incorporating hierarchical Transformer blocks and an attention mechanism to enhance feature representation capabilities. Secondly, to alleviate inter-domain radiometric discrepancies and attain image-level alignment, a Wallis filter is utilized to perform image style transformation. Considering the class imbalance present in the landslide dataset, a Rare Class Sampling strategy is introduced to mitigate bias towards common classes and strengthen the learning of the rare landslide class. Finally, a contrastive loss is adopted to further optimize and enhance the model’s ability to delineate fine-grained class boundaries. The proposed model is validated on the Potsdam and Vaihingen benchmark datasets, followed by validation in two landslide scenarios induced by earthquakes and rainfall to evaluate its adaptability across different disaster domains. Compared to the source-only model, LandsDANet achieved improvements in IoU of 27.04% and 35.73% in two cross-domain landslide disaster recognition tasks, respectively. This performance not only showcases its outstanding capabilities but also underscores its robust potential to meet the demands for rapid response. Full article

Electroencephalography (EEG)-based motor imagery (MI) brain–computer interfaces (BCIs) hold considerable potential for applications in neuro-rehabilitation and assistive technologies. Yet, their development remains constrained by challenges such as low spatial resolution, vulnerability to noise and artifacts, and pronounced inter-subject variability. Conventional approaches, including common spatial patterns (CSP) and convolutional neural networks (CNNs), often exhibit limited robustness, weak generalization, and reduced interpretability. To overcome these limitations, we introduce EEG-GCIRNet, a Gaussian connectivity-driven EEG imaging representation network coupled with a regularized LeNet architecture for MI classification. Our method integrates raw EEG signals with topographic maps derived from functional connectivity into a unified variational autoencoder framework. The network is trained with a multi-objective loss that jointly optimizes reconstruction fidelity, classification accuracy, and latent space regularization. The model’s interpretability is enhanced through its variational autoencoder design, allowing for qualitative validation of its learned representations. Experimental evaluations demonstrate that EEG-GCIRNet outperforms state-of-the-art methods, achieving the highest average accuracy (81.82%) and lowest variability (±10.15) in binary classification. Most notably, it effectively mitigates BCI illiteracy by completely eliminating the “Bad” performance group (<60% accuracy), yielding substantial gains of ∼22% for these challenging users. Furthermore, the framework demonstrates good scalability in complex 5-class scenarios, performing competitive classification accuracy (75.20% ± 4.63) with notable statistical superiority (p = 0.002) against advanced baselines. Extensive interpretability analyses, including analysis of the reconstructed connectivity maps, latent space visualizations, Grad-CAM++ and functional connectivity patterns, confirm that the model captures genuine neurophysiological mechanisms, correctly identifying integrated fronto-centro-parietal networks in high performers and compensatory midline circuits in mid-performers. These findings suggest that EEG-GCIRNet provides a robust and interpretable end-to-end framework for EEG-based BCIs, advancing the development of reliable neurotechnology for rehabilitation and assistive applications. Full article

Background: Cancer poses a significant health risk to humans, with prostate cancer (PCa) being the second most common and deadly form among men, following lung cancer. Each year, it affects over a million individuals and presents substantial diagnostic challenges due to variations in tissue appearance and imaging quality. In recent decades, various techniques utilizing Magnetic Resonance Imaging (MRI) have been developed for identifying and classifying PCa. Accurate classification in MRI typically requires the integration of complementary feature types, such as deep semantic representations from Convolutional Neural Networks (CNNs) and handcrafted descriptors like Histogram of Oriented Gradients (HOG). Therefore, a more robust and discriminative feature integration strategy is crucial for enhancing computer-aided diagnosis performance. Objectives: This study aims to develop a multi-stage hybrid learning model that combines deep and handcrafted features, investigates various feature reduction and classification techniques, and improves diagnostic accuracy for prostate cancer using magnetic resonance imaging. Methods: The proposed framework integrates deep features extracted from convolutional architectures with handcrafted texture descriptors to capture both semantic and structural information. Multiple dimensionality reduction methods, including singular value decomposition (SVD), were evaluated to optimize the fused feature space. Several machine learning (ML) classifiers were benchmarked to identify the most effective diagnostic configuration. The overall framework was validated using k-fold cross-validation to ensure reliability and minimize evaluation bias. Results: Experimental results on the Transverse Plane Prostate (TPP) dataset for binary classification tasks showed that the hybrid model significantly outperformed individual deep or handcrafted approaches, achieving superior accuracy of 99.74%, specificity of 99.87%, precision of 99.87%, sensitivity of 99.61%, and F1-score of 99.74%. Conclusions: By combining complementary feature extraction, dimensionality reduction, and optimized classification, the proposed model offers a reliable and generalizable solution for prostate cancer diagnosis and demonstrates strong potential for integration into intelligent clinical decision-support systems. Full article

Deep learning methods based on convolutional neural networks (CNNs) and Mamba have advanced medical image segmentation, yet two challenges remain: (1) trade-off in feature extraction, where CNNs capture local details but miss global context, and Mamba captures global dependencies but overlooks fine structures, and (2) limited feature aggregation, as existing methods insufficiently integrate inter-layer common information and delta details, hindering robustness to subtle structures. To address these issues, we propose a hybrid cascade and dual-path adaptive aggregation network (HCDAA-Net). For feature extraction, we design a hybrid cascade structure (HCS) that alternately applies ResNet and Mamba modules, achieving a spatial balance between local detail preservation and global semantic modeling. We further employ a general channel-crossing attention mechanism to enhance feature expression, complementing this spatial modeling and accelerating convergence. For feature aggregation, we first propose correlation-aware aggregation (CAA) to model correlations among features of the same lesions or anatomical structures. Second, we develop a dual-path adaptive feature aggregation (DAFA) module: the common path captures stable cross-layer semantics and suppresses redundancy, while the delta path emphasizes subtle differences to strengthen the model’s sensitivity to fine details. Finally, we introduce a residual-gated visual state space module (RG-VSS), which dynamically modulates information flow via a convolution-enhanced residual gating mechanism to refine fused representations. Experiments on diverse datasets demonstrate that our HCDAA-Net outperforms some state-of-the-art (SOTA) approaches. Full article

This article compares the roles of allotments and community gardens in democratising London’s urban food system. Drawing from ethnographic and participatory action research (PAR), it reveals a recent policy shift favouring community gardens compared to allotments, which has resulted in a net reduction in long-term urban agriculture space in London. The study contrasts these two trajectories of urban agriculture across five democratic processes: (1) fostering food security, (2) expanding health benefits, (3) reclaiming the commons, (4) building spaces of interaction and representation, and (5) decoupling from dominant regimes. While community gardens tend to perform well in terms of social inclusion and environmental education of local communities and marginalised populations, allotments tend to be more successful in terms of productive capacity and developing autonomy due to their relatively more secure tenure. However, both trajectories are increasingly challenged by the dynamics of neoliberal urban development and the withdrawal of the state from its welfare responsibilities. This article argues that both trajectories do not have to be mutually exclusive and that their coexistence is in fact necessary to develop a more resilient urban food system, one that realises the principles of food sovereignty, social justice, and agroecological urbanisms at the local level. Full article

In the classification applications of hyperspectral remote sensing images (HSIs), band selection is crucial for mitigating the curse of dimensionality while preserving the intrinsic physical information within HSIs. Although clustering-based band selection methods are widely applied, they often overlook the inherent physical properties of hyperspectral images. Such approaches typically operate in raw high-dimensional space, which is susceptible to noise and redundancy. This results in generated band combinations that fail to adequately characterize the spectral features of the underlying materials, leading to suboptimal band-grouping schemes. To address this, we propose a novel Interband Consistency-Constrained Structural Subspace Clustering (ICC-SSC) method. The core assumption is that the spectral characteristics of land cover inherently reside within a low-dimensional subspace, where bands within this subspace should exhibit strong physical consistency, which means that the spectral signatures of land covers show significant similarity across these bands. Driven by this physical interpretation, our method innovates in two ways. Specifically, we employ the $l_{1,2}$ norm in the self-representation model to discover the inherent grouping structure of the bands. This enforces a small set of common, representative basis bands to reconstruct others, effectively identifying the most physically informative bands that anchor these material-specific subspaces. In addition, we incorporate a total variance (TV) regularization term into the proposed model to capture the smoothing characteristics between adjacent bands. This physics-based constraint enhances the consistency of representations among adjacent bands, ensuring that subspace representations across all bands maintain well-structured coherence. An efficient algorithm based on the Alternating Direction Method of Multipliers (ADMM) is derived to solve the proposed model. Extensive experiments on three real HSIs demonstrate that ICC-SSC significantly outperforms state-of-the-art methods. Full article

Biplot methods provide a framework for the simultaneous graphical representation of both rows and columns of a data matrix. Classical biplots were originally developed for continuous data in conjunction with principal component analysis (PCA). In recent years, several extensions have been proposed for binary and nominal data. These variants, referred to as logistic biplots (LBs), are based on logistic rather than linear response models. However, existing formulations remain insufficient for analyzing ordinal data, which are common in many social and behavioral research contexts. In this study, we extend the biplot methodology to ordinal data and introduce the ordinal logistic biplot (OLB). The proposed method estimates row scores that generate ordinal logistic responses along latent dimensions, whereas column parameters define logistic response surfaces. When these surfaces are projected onto the space defined by the row scores, they form a linear biplot representation. The model is based on a framework, leading to a multidimensional structure analogous to the graded response model used in Item Response Theory (IRT). We further examine the geometric properties of this representation and develop computational algorithms—based on an alternating gradient descent procedure—for parameter estimation and computation of prediction directions to facilitate visualization. The OLB method can be viewed as an extension of multidimensional IRT models, incorporating a graphical representation that enhances interpretability and exploratory power. Its primary goal is to reveal meaningful patterns and relationships within ordinal datasets. To illustrate its usefulness, we apply the methodology to the analysis of job satisfaction among PhD holders in Spain. The results reveal two dominant latent dimensions: one associated with intellectual satisfaction and another related to job-related aspects such as salary and benefits. Comparative analyses with alternative techniques indicate that the proposed approach achieves superior discriminatory power across variables. Full article

Background: Accurately predicting protein druggability is crucial for successful drug development, as it significantly reduces the time and resources required to identify viable drug targets. However, existing methods often face trade-offs between accuracy, efficiency, and interpretability. This study aims to introduce a lightweight framework designed to address these challenges effectively. Methods: We present a lightweight framework that embeds proteins into four biologically informed, non-Euclidean metric spaces, derived from analyses of amino acid sequences, predicted secondary structures, and curated post-translational modification (PTM) annotations. These representations capture key features such as hydrophobicity profiles, PTM densities, spatial patterns, and secondary structure composition, providing interpretable proxies for structure-related determinants of druggability. This approach enhances our understanding of protein functionality while improving druggability predictability in a biologically relevant context. Results: Evaluated on an Aspirin-binding protein dataset using leave-one-out cross-validation (LOOCV), our distance-based ensemble achieves 92.25% accuracy (AUC = 0.9358) in the whole-protein setting. This performance significantly outperforms common sequence-only baselines in the literature while remaining computationally efficient. Conclusions: On a refined single-chain subset, our framework demonstrates performance comparable to established feature engineering pipelines, highlighting its potential effectiveness in practical applications. Together, these results strongly suggest that biologically grounded, non-Euclidean embeddings provide an effective and interpretable alternative to resource-intensive 3D pipelines for target assessment in drug discovery. This approach not only enhances our ability to assess protein druggability but also streamlines the overall process of target identification and validation. Full article

Zero-shot learning (ZSL) aims to categorize target classes with the aid of semantic knowledge and samples from previously seen classes. In this process, the alignment of visual and attribute modality features is key to successful knowledge transfer. Several previous studies have investigated the extraction of attribute-related local features to reduce visual-semantic domain gaps and overcome issues with domain shifts. However, these techniques do not emphasize the commonality of features across different objects belonging to the same attribute, which is critical for identifying and distinguishing the attributes of unseen classes. In this study, we propose a novel ZSL method, termed dual-contrastive attribute embedding (DCAE), for generalized zero-shot learning. This approach simultaneously learns both class-level and attribute-level prototypes and representations. Specifically, an attribute embedding module is introduced to capture attribute-level features and an attribute semantic encoder is developed to generate attribute prototypes. Attribute-level and class-level contrastive loss terms are then used to optimize an attribute embedding space such that attribute features are compactly distributed around corresponding prototypes. This double contrastive learning mechanism facilitates the alignment of multimodal information from two dimensions. Extensive experiments with three benchmark datasets demonstrated the superiority of the proposed method compared to current state-of-the-art techniques. Full article

Background: Ethnic minority groups are under-represented in their use of community mental health services in the UK. Online mental health forums could be a more appealing support option than traditional mental health services. Part one of this study investigated the level of online forum use in people from ethnic minority groups. Part two investigated the factors influencing online mental health forum use for people from ethnic minority groups. Methods: Part one involved comparing data from a range of pre-existing national datasets, and datasets local to Berkshire, UK (i.e., on the general population, people experiencing common mental health problems, users of mental health forums, and NHS Talking Therapies services). Part two involved interviewing 14 individuals from ethnic minority backgrounds who had used, or considered using, online mental health forums. Results: In part one, nationally, Asian, Black, and Mixed ethnic groups appeared over-represented in their online mental health forum use based on their reporting of common mental health problems. In Berkshire, people from Asian and Black ethnic groups were under-represented in their use of Berkshire NHS Trust’s online mental health forum based on their representation in the Berkshire population. In Part Two, three themes were identified as influencing forum use: (1) sense of community in the online and offline worlds, (2) trust is crucial, and (3) barriers to accessing online forums. Conclusion: People from ethnic minority groups vary in their use and experiences of mental health forums. Whilst forums can offer a valued accessible space for anonymous sharing of often stigmatised experiences, pathways to access require trusted figures to promote their availability, and forum designers and moderators to co-create culturally sensitive spaces with people from these target communities. Full article

Unmanned Aerial Vehicles (UAVs) are increasingly integral to construction site management, supporting monitoring, inspection, and data collection tasks. Effective UAV path planning is essential for maximizing operational efficiency, particularly in complex and dynamic construction environments. While previous BIM-based approaches have explored representation models such as space graphs, grid patterns, and voxel models, each has limitations. Space graphs, though common, rely on predefined spatial spaces, making them less suitable for projects still under construction. Voxel-based methods, considered well-suited for 3D indoor navigation, suffer from three key challenges: (1) a disconnect between the BIM and voxel models, limiting data integration; (2) the computational cost and time required for voxelization, hindering real-time application; and (3) inadequate support for 4D BIM integration during active construction phases. This research introduces a novel framework that bridges the BIM–voxel gap via an enriched voxel map, eliminates the need for repeated voxelization, and incorporates 4D BIM and additional model data such as defined workspaces and safety buffers around fragile components. The framework’s effectiveness is demonstrated through path planning simulations on BIM models from two real-world construction projects under varying scenarios. Results indicate that the enriched voxel map successfully creates a connection between BIM model and voxel model, while covering every timestamp of the project and element attributes during path planning without requiring additional voxel map creation. Full article