Search Results (7,874)

Polarimetric descattering imaging has attracted growing interest due to its fundamental physical significance and potential applications. While deep learning has accelerated its development through powerful feature extraction and inference capabilities, existing methods still face limitations in practical scenarios, particularly under dynamic non-uniform scattering conditions such as cement dust environments. To address this, we propose a deep neural network based on the Mueller matrix model that effectively integrates polarization evolution information with deep learning. Specifically, local concentrations of the scattering medium in non-uniform cement dust are characterized by the evolution of the degree of linear polarization (DoLP), which is converted into pixel-wise weight biases to generate customized Mueller matrices adaptable to varying concentrations. The network predicts a pixel-wise dust concentration map and applies the corresponding concentration-specific Mueller matrix to each pixel for polarization-aware dehazing, ensuring physical consistency with Mueller matrix calculus throughout inference. This framework is further enhanced by a physics-constrained optimization loss and multi-scale feature fusion. Experimental results demonstrate the method’s effectiveness and superiority in diverse dynamic non-uniform cement dust environments. Full article

The rapid advancements in large language models (LLMs) have transformed natural language processing, enabling their application in diverse domains such as conversational agents and decision-support systems in sensitive areas like healthcare, finance, and eldercare. However, as LLMs are increasingly integrated into real-world contexts, concerns about their adherence to ethical principles, privacy norms, and contextual expectations have become critical. Privacy preservation is particularly pressing in interactions involving personal or sensitive data, where ensuring that LLMs align with societal norms while mitigating risks of information leakage is essential to fostering trust and ensuring responsible deployment. Contextual integrity (CI) provides a robust framework to address these challenges, emphasizing that information flows should adhere to context-specific social norms. This principle is especially vital in sensitive applications, where LLMs must evaluate roles, information attributes, and transmission principles to maintain ethical behavior. Despite their linguistic proficiency, LLMs often fail to recognize and adapt to nuanced contextual norms, a limitation exacerbated by their probabilistic nature and the biases in their training data, which can lead to inappropriate or harmful outputs. Addressing these shortcomings requires rigorous evaluation methodologies and fine-tuning strategies that embed societal and contextual norms into the models. Full article

Ultra-wideband (UWB) technology has been widely adopted for indoor positioning due to its high temporal resolution. However, the accuracy of UWB-based indoor positioning is fundamentally limited by ranging measurement errors, particularly under non-line-of-sight (NLOS) conditions, where systematic bias and uncertainty are introduced into the measured distances. In this paper, a measurement error mitigation method is proposed to improve UWB ranging reliability in complex indoor environments. The method first identifies NLOS measurements using low-dimensional physical features and a lightweight machine learning classifier. Subsequently, an error compensation strategy is applied to correct biased ranging observations, which are then incorporated into a nonlinear least squares positioning model. Experimental results obtained in typical indoor environments demonstrate that the proposed method significantly reduces ranging errors and improves positioning accuracy compared with conventional approaches. The results indicate that the proposed framework effectively enhances measurement robustness without increasing system complexity. Full article

We present the performance of a middle atmosphere Rayleigh lidar system that employs three receiver channels. We characterize the biases in the density and temperature profiles retrieved from each of the receiver channels as well as the combined receiver signal. We associate these biases with pulse pile-up, gain switching, and variations in the detector gain due to signal amplitude. We use a top-down temperature convergence methodology to determine the upper altitude up to which the signals should be compensated for the variations in detector gain. We find that the channels have warm biases in their temperatures of 2–8 K at 40 km. These biases decrease to between 1 K and 3 K at 60 km. Uncertainty estimates derived from the photon-counting statistics indicate temperature uncertainties on the order of 2–5 K in the 40–70 km region, which are consistent with the observed level of inter-channel variability after correction. A comparison with MERRA-2 reanalysis indicates an overall agreement in temperatures and differences that are consistent with the comparisons between the Rayleigh lidars and MERRA-02 at other sites. These results demonstrate that the proposed approach proves reliable for processing the multi-channel Rayleigh lidar data, particularly for systems employing more than two detection channels, and improves the fidelity and accuracy of the temperature retrievals. Full article

Background/Objectives: Breast cancer continues to be one of the most serious and common afflictions affecting women around the globe. Despite ultrasound imaging being an effective method for the detection of abnormalities in dense breast tissues, there are a number of drawbacks when utilizing this method, including the subjective nature of the imaging and the variant nature of the imaging due to the cognitive biases of the interpreting expert and the experience of the interpreting expert. The above factors are the cause of the increased need in the implementation of AI-driven models for diagnostic analysis. In this research, we provide a hybrid deep learning-based framework for cancer classification of the breast cancer ultrasound image dataset (‘BUSI dataset’). Methods: The contributing models of the proposed architecture involve the combination of a light ViT encoder and an EfficientNetV2-RW-S feature extractor. The combination mentioned leverage the positive sensitivities of the convolutional neural networks (CNNs) and the global reasoning neural networks (i.e., transformers) in the explanation of the architecture. The reason being, EfficientNetV2 diminishes the capture of the fine-grained morphological components of the lesions, edges, and echogenic variances of the tissue, whereas the transformer model diminishes the long-range dependencies of the lesions and other surrounding tissues. Results: The experimental results from the proposed hybrid model of the architecture demonstrates an enhanced classification accuracy of 97.95%, in contrast to the self-standing models of the architecture, the hybrid model supersedes the isolated ViT model (i.e., 89%) and the isolated CNN model (i.e., 80%) frameworks. Furthermore, the proposed model hybrid architecture also diminishes the overall self-attention computational complexity of the proposed model by substantially diminishing the number of tokens reaching an overall count of 10 (from the vast 197 tokens). This further leads to a substantial decrease in the memory and cost expended during the attention processes. Conclusion: Overall, this study proposes a method for the improved diagnostic and computational analysis, suggesting the proposed architecture to be a potential framework for use in the contemporary clinical environments. Full article

Background: Generative artificial intelligence (AI) can produce realistic human faces that are shared on social media, from where younger generations often derive body image norms. Aesthetic bias in these systems may promote unrealistic standards of beauty. This study examines whether generative AI produces facial images that are perceived by humans as more attractive than real human faces. Thus, we examine AI-generated facial imagery as a contemporary site of consumer culture, where beauty may become biased, unrealistic, and commodified: generating an algorithmically optimized product circulating through social media and digital platforms without proper regulation. Methods: Fifty AI-generated female faces were prospectively compared with 50 photographs of female models from a model agency. Facial attractiveness was rated by plastic surgeons, using a Likert scale and Mann–Whitney U for analysis. Results: AI-generated images received higher mean aesthetic scores than real photographs (7.79 vs. 6.88, p < 0.05), despite prompts requesting unattractive features. Conclusions: The AI model showed a small but consistent bias toward enhanced facial attractiveness. As AI-generated imagery increasingly shapes visual culture, this bias may contribute to unrealistic beauty standards, highlighting the need for AI literacy, responsible use of AI, and ethical oversight, especially when shared on social media. Full article

Accurate identification of fly species composition and their responses to attractants is critical for risk assessment and targeted vector control. To evaluate the efficacy of different attractants in surveillance and their species-specific trapping biases, a standardized field study was conducted from June to September 2021 across seven representative cities in China’s major zoogeographical regions: Xining, Ürümqi, Yanji, Beijing, Chongqing, Kunming, and Sanya. Cage traps baited with either fish offal or sugar–vinegar solution were deployed, supplemented by hand-net collection. A total of 134 traps were set, yielding 2132 flies belonging to 21 species. Fish offal captured 1961 flies (91.9%), significantly more than the 101 flies (4.7%) caught with sugar–vinegar solution (χ² = 1582.3, p < 0.001). Lucilia sericata was the dominant species (885 individuals, 41.51%), followed by L. cuprina (178, 8.35%), Sarcophaga portschinskyi (127, 5.96%), and Sarcophaga africa (100, 4.70%). High-risk taxa (Calliphoridae and Sarcophagidae) were almost exclusively attracted to fish offal. Our findings demonstrate that protein-based baits, such as fish offal, are substantially more effective than traditional sugar–vinegar solutions for capturing epidemiologically relevant fly species across diverse ecological zones in China. We recommend prioritizing proteinaceous attractants in national fly surveillance programs and advocate for routine species-level identification to enable risk-informed vector monitoring. Full article

Hematopoietic stem and progenitor cells (HSPCs) in the bone marrow are highly vulnerable to radiation-induced damage. Systematic delineation of lineage-specific transcription factor (TF) programs, together with in silico perturbation analyses, provides a valuable approach for identifying regulators capable of accelerating hematopoietic reconstruction after irradiation. Here, using single-cell RNA sequencing (scRNA-seq), we characterized the dynamics of HSPCs at both cellular abundance and transcriptional regulation levels following irradiation and used in silico TF perturbation to predict their effects on lineage commitment. We found that granulocyte–macrophage progenitor (GMP) differentiation is consistently prioritized after irradiation, accompanied by enhanced activity of proliferation-associated drivers. Network-based TF profiling identified Tcf7l2 as a previously unrecognized regulator of early lymphoid differentiation. In silico perturbation further functionally predicted TFs driving differentiation in HSPCs after irradiation, and Hsf1, a factor with pharmacological activation potential, was selected for validation via in vivo celastrol treatment and in vitro knockdown. Collectively, our findings uncover the transcriptional programs governing HSPC lineage biases after radiation exposure and highlight the utility of in silico TF perturbation as a strategy for guiding the therapeutic interventions for radiation-induced hematopoietic injury. Full article

Solving constrained multiobjective optimization problems (CMOPs) is highly challenging due to the presence of complicated feasible regions, intense conflicts among objectives, and unevenly distributed constraints. As a result, conventional methods relying on a single constraint-handling mechanism frequently fail to maintain a stable equilibrium among solution feasibility, diversity, and convergence. To overcome these bottlenecks, this article introduces AFFCMO, a novel adaptive feasibility-guided framework tailored for constrained multiobjective optimization. At its core, the proposed approach utilizes a coevolutionary dual-population architecture that divides the search process into two distinct tasks. Specifically, an auxiliary population is tasked with global exploration, while a primary population focuses on the intensive exploitation of discovered feasible areas. To achieve this, the primary population leverages a DE/current-to-pbest/1 differential evolution strategy to closely approximate the constrained Pareto front. Simultaneously, the auxiliary population expands the search space using a mutation operator that adapts to the current evolutionary stage. Furthermore, exploration is bolstered by a multicriterion environmental selection scheme designed for the auxiliary group. By combining Euclidean geometric distributions, constraint relaxation, and value modeling inspired by epidemic dynamics, this strategy successfully preserves valuable infeasible solutions that can guide the search. Additionally, a dynamic resource allocation strategy based on historical search feedback and Thompson sampling is incorporated. This mechanism continuously evaluates the recent search contributions of both populations and adaptively adjusts their offspring sizes, thereby reducing the bias introduced by static allocation schemes. This mechanism continuously assesses the actual search contributions of both populations, allowing for the adaptive resizing of offspring generations and thereby eliminating the inherent biases of static allocation methods. Comprehensive empirical evaluations are conducted on 47 benchmark problems from four distinct test suites. The results indicate that AFFCMO significantly outperforms seven contemporary multiobjective evolutionary algorithms in terms of exploring complex feasible regions, preserving solution diversity, and achieving high convergence accuracy. Full article

Contemporary Music Information Retrieval (MIR) and Natural Language Processing (NLP) systems are increasingly applied to diverse musical traditions, yet they are largely grounded in Western musical and linguistic assumptions. This study examines whether commonly used MIR features and multilingual NLP models adequately represent the acoustic, linguistic, and cultural structures of Maskandi music in comparison to Western music and identifies where representational gaps and biases arise. A multidimensional framework was employed, comprising acoustic and structural MIR analysis, linguistic and semantic lyrical analysis, and bias analysis. A curated dataset of 60 recordings and corresponding lyrics was analysed using rhythm and beat features, pitch contour measures, structural self-similarity, timbre embeddings, semantic similarity, lexical diversity, metaphor density, topic modelling, multilingual embeddings, and dataset-level audits. The results reveal systematic representational failures: beat tracking showed lower median IOI coefficient of variation for Maskandi (0.028) versus Western music (0.040, p = 0.0199) yet exhibited greater algorithmic instability, tempo averaged 131.16 BPM versus 111.69 BPM (p = 0.000262), pitch glide proportions were significantly higher in Maskandi (0.34 vs. 0.16), on-beat energy ratios differed substantially (2.26 vs. 1.19, p < 0.0000007), semantic similarity revealed high intra-genre coherence for Maskandi (0.73) versus Western (0.25), metaphor density approached zero in Maskandi versus up to 7 per 100 words in Western lyrics, topic modeling produced two compact clusters for Maskandi versus 6 dispersed clusters for Western, timbre embeddings achieved a 0.405 silhouette score, dataset audits revealed 0% Maskandi representation across seven major MIR corpora with African traditions comprising <3%. The study concludes that statistical separability does not imply representational adequacy and highlights the need for culturally grounded MIR and NLP representations to support diverse musical traditions. Full article

Background: Neurodegenerative diseases are a major health problem, and the neuropsychiatric symptoms seen in these diseases adversely impact the lives of patients, families, and caregivers. Inappropriate aggressive behavior is a highly disruptive symptom and a leading cause of institutionalization. There are no approved drugs specifically for the treatment of problematic aggression, and the off-label use of antipsychotics has limited benefit with significant side effects and safety risks. This review discusses dysregulated arginine vasopressin (AVP) signaling in fear–threat circuitry as a key driver of inappropriate aggression. Because the AVP 1a receptor (V1aR) is the dominant subtype in the CNS, the selective antagonism of this receptor represents a well-rationalized target for the treatment of aggression across neurodegenerative, psychiatric, and neurodevelopmental disorders. Objectives: Our goal was to summarize the basis for using V1aR antagonists as a treatment for irritability and aggressive behavior. We describe its discovery, biosynthesis, receptor pharmacology, and CNS distribution, emphasizing V1aR localization in central fear–threat circuits. Translational evidence from animal studies, pharmacological neuroimaging, and lesion network mapping is presented. These data support the suggestion that heightened vasopressinergic tone biases socioemotional information processing toward negative valence, increasing threat sensitivity and the likelihood of inappropriate aggressive responses. Emerging clinical data support this framework. Highly selective, CNS-penetrant V1aR antagonists reduced aggressive behavior and had an excellent safety profile in phase 2 studies in Huntington’s disease and intermittent explosive disorder, with efficacy signals across caregiver-reported, clinician-rated, and incident-based measures. Furthermore, pharmacological neuroimaging showed that V1aR antagonism normalizes AVP-induced alterations in activity within fear–threat circuitry. Conclusions and Future Directions: Preclinical, translational, and clinical findings to date support V1aR antagonism as a promising strategy for treating pathological aggression across disorders. Additional experimental medicine studies and clinical trials are needed to conclusively establish efficacy in various disease populations, and we note the need for improved trial designs and analytical methods as part of the development process. Full article

This study analyzes the relationship between stock market returns and investment flows in investment funds in Spain. Through a quantitative analysis covering the period from December 2001 to June 2025, it examines not only the existence of a correlation but also its temporal structure, functional form, and heterogeneity across different geographical areas (U.S., Europe, Japan, and Spain). Using monthly data on net flows from INVERCO and market indices, the study employs Ordinary Least Squares (OLS) regression models, segmented regressions, and fixed-effects panel models to obtain robust estimates. The results confirm a positive and statistically significant relationship between past returns and subsequent investment flows, with a temporal lag ranging from one to three months. This delay varies notably by geographical region, suggesting the existence of different investor profiles and information channels. The study also finds evidence of a convex relationship, indicating that investors react asymmetrically, aggressively pursuing high returns more than penalizing low ones. These findings, interpreted through the lens of behavioral finance, point to pro-cyclical and reactive behavior of Spanish investors, driven by biases such as loss aversion, trend-following, and delays in information processing. The study contributes to the academic literature by providing updated and methodologically robust evidence on Spain, a market that has traditionally been underexplored, and offers practical implications for investors, fund managers, and regulators in terms of financial education and risk management. Full article

The traditional paradigm of pharmaceutical research is characterized by substantial inefficiency, requiring extensive timelines and billions of dollars while suffering from high clinical attrition rates. The integration of generative artificial intelligence (AI) is driving a paradigm shift from empirical experimentation toward predictive, data-driven innovation. This review evaluates state-of-the-art applications of these technologies across the drug discovery and development pipeline. By analyzing multi-omics data streams, AI models can elucidate complex disease mechanisms and identify novel therapeutic targets. Deep generative architectures facilitate the algorithmic creation of novel molecular entities, enabling the design of therapeutics with complex polypharmacological profiles. Furthermore, AI is enhancing the clinical testing phase through large language models (LLMs) that improve patient enrollment and through synthetic control arms (SCAs) that provide computational alternatives to traditional placebo groups. Despite these advances, the scientific community must address inherent algorithmic biases stemming from demographic underrepresentation and mitigate the risks of data hallucinations. Ultimately, realizing the full translational potential of generative AI in precision medicine may require the widespread adoption of explainable AI (XAI) frameworks and rigorous data standards. Full article

In this article, a low-power low-intermediate-frequency (Low-IF) receiver front-end is presented for Bluetooth Low Energy (BLE) body area network (BAN) applications. The receiver employs an input matching network, an inductorless self-biased inverter-based low-noise transconductance amplifier (LNTA), a single-balanced passive mixer, a common-gate transimpedance amplifier (TIA), and a Gm-C complex filter for image suppression. Native MOS devices are adopted to support low-voltage operation and reduce static power consumption. The interstage on-chip coupling capacitor between the RF front-end and the TIA is removed by aligning the DC operating points of the two stages. The receiver front-end is implemented in a 55 nm standard CMOS process and occupies an active area of 0.081 ${\mathrm{mm}}^2$, excluding bonding pads. Post-layout simulations show that the proposed design achieves 45.2 dB gain, 7.2 dB noise figure, and 28.1 dB image rejection ratio over the 2.4–2.48 GHz band, while consuming 537 $\mathsf{\mu }$W. The proposed front-end is suitable for low-power BLE BAN sensor nodes. Full article

Scene text detection aims to accurately localize text instances in images captured under complex environments. Its performance depends heavily on precise text boundary delineation and reliable semantic discrimination from cluttered backgrounds. However, existing methods still struggle in such complex scenes. Repeated downsampling gradually biases features toward low-frequency components, thereby weakening edge details and local structures that are critical to text morphology. Additionally, semantic information and local details are often modeled independently. This lack of coordination makes high-frequency responses vulnerable to background noise. To address these issues, we propose HFI-Former, a Transformer-based model designed for high-frequency enhancement and feature interaction. The framework consists of multi-scale feature extraction, frequency-enhanced representation, semantic-guided feature interaction, and deformable Transformer encoding. Frequency-domain enhancement is introduced to preserve high-frequency structural features degraded by repeated downsampling. Semantic-aware feature interaction further injects global context to regulate multi-scale feature fusion. Experiments on CTW1500, Total-Text and ICDAR1500 demonstrate competitive boundary localization accuracy and strong overall detection performance in complex scenes. Full article