Search Results (22,944)

Adversarial attacks in Natural Language Processing (NLP) present a critical challenge, particularly in sentiment analysis, where subtle input modifications can significantly alter model predictions. In search of more robust defenses against adversarial attacks on sentimental analysis, this research work introduces two novel defense mechanisms: the Lexicon-Based Random Substitute Model (LRSM) and the Word-Variant Voting Model (WVVM). LRSM employs randomized substitutions from a dataset-specific lexicon to generate diverse input variations, disrupting adversarial strategies by introducing unpredictability. Unlike traditional defenses requiring synonym dictionaries or precomputed semantic relationships, LRSM directly substitutes words with random lexicon alternatives, reducing overhead while maintaining robustness. Notably, LRSM not only neutralizes adversarial perturbations but occasionally surpasses the original accuracy by correcting inherent model misclassifications. Building on LRSM, WVVM integrates LRSM, Frequency-Guided Word Substitution (FGWS), and Synonym Random Substitution and Voting (RS&V) in an ensemble framework that adaptively combines their outputs. Logistic Regression (LR) emerged as the optimal ensemble configuration, leveraging its regularization parameters to balance the contributions of individual defenses. WVVM consistently outperformed standalone defenses, demonstrating superior restored accuracy and F1 scores across adversarial scenarios. The proposed defenses were evaluated on two well-known sentiment analysis benchmarks: the IMDB Sentiment Dataset and the Yelp Polarity Dataset. The IMDB dataset, comprising 50,000 labeled movie reviews, and the Yelp Polarity dataset, containing labeled business reviews, provided diverse linguistic challenges for assessing adversarial robustness. Both datasets were tested using 4000 adversarial examples generated by established attacks, including Probability Weighted Word Saliency, TextFooler, and BERT-based Adversarial Examples. WVVM and LRSM demonstrated superior performance in restoring accuracy and F1 scores across both datasets, with WVVM excelling through its ensemble learning framework. LRSM improved restored accuracy from 75.66% to 83.7% when compared to the second-best individual model, RS&V, while the Support Vector Classifier WVVM variation further improved restored accuracy to 93.17%. Logistic Regression WVVM achieved an F1 score of 86.26% compared to 76.80% for RS&V. These findings establish LRSM and WVVM as robust frameworks for defending against adversarial text attacks in sentiment analysis. Full article

I address the challenge of extracting reliable insights from large datasets using a simplified model that illustrates how hierarchical classification can distort outcomes. The model consists of discrete pixels labeled red, blue, or white. Red and blue indicate distinct properties, while white represents unclassified or ambiguous data. A macro-color is assigned only if one color holds a strict majority among the pixels. Otherwise, the aggregate is labeled white, reflecting uncertainty. This setup mimics a percolation threshold at fifty percent. Assuming that directly accessing the various proportions from the data of colors is infeasible, I implement a hierarchical coarse-graining procedure. Elements (first pixels, then aggregates) are recursively grouped and reclassified via local majority rules, ultimately producing a single super-aggregate for which the color represents the inferred macro-property of the collection of pixels as a whole. Analytical results supported by simulations show that the process introduces additional white aggregates beyond white pixels, which could be present initially; these arise from groups lacking a clear majority, requiring arbitrary symmetry-breaking decisions to attribute a color to them. While each local resolution may appear minor and inconsequential, their repetitions introduce a growing systematic bias. Even with complete data, unavoidable asymmetries in local rules are shown to skew outcomes. This study highlights a critical limitation of recursive data reduction. Insight extraction is shaped not only by data quality but also by how local ambiguity is handled, resulting in built-in biases. Thus, the related flaws are not due to the data but to structural choices made during local aggregations. Although based on a simple model, these findings expose a high likelihood of inherent flaws in widely used hierarchical classification techniques. Full article

The advancement of Artificial Intelligence (AI) has significantly accelerated progress across various research domains, with growing interest in plant science due to its substantial economic potential. However, the integration of AI with digital vegetation analysis remains underexplored, largely due to the absence of large-scale, real-world plant datasets, which are crucial for advancing this field. To address this gap, we introduce the PP3D dataset—a meticulously labeled collection of about 500 potted plants represented as 3D point clouds, featuring fine-grained annotations for approximately 20 species. The PP3D dataset provides 3D phenotypic data for about 20 plant species spanning model organisms (e.g., Arabidopsis thaliana), potted plants (e.g., Foliage plants, Flowering plants), and horticultural plants (e.g., Solanum lycopersicum), covering most of the common important plant species. Leveraging this dataset, we propose the panoptic plant recognition task, which combines semantic segmentation (stems and leaves) with leaf instance segmentation. To tackle this challenge, we present SCNet, a novel dual-representation learning network designed specifically for plant point cloud segmentation. SCNet integrates two key branches: a cylindrical feature extraction branch for robust spatial encoding and a sequential slice feature extraction branch for detailed structural analysis. By efficiently propagating features between these representations, SCNet achieves superior flexibility and computational efficiency, establishing a new baseline for panoptic plant recognition and paving the way for future AI-driven research in plant science. Full article

To address the inefficiency of traditional concrete crack detection methods and the heavy reliance of supervised learning on extensive labeled data, in this study, an intelligent assessment method of concrete damage based on pseudo-label semi-supervised learning and fractal geometry theory is proposed to solve two core tasks: one is binary classification of pixel-level cracks, and the other is multi-category assessment of damage state based on crack morphology. Using three-channel RGB images as input, a dual-path collaborative training framework based on U-Net encoder–decoder architecture is constructed, and a binary segmentation mask of the same size is output to achieve the accurate segmentation of cracks at the pixel level. By constructing a dual-path collaborative training framework and employing a dynamic pseudo-label refinement mechanism, the model achieves an F1-score of 0.883 using only 50% labeled data—a mere 1.3% decrease compared to the fully supervised benchmark DeepCrack (F1 = 0.896)—while reducing manual annotation costs by over 60%. Furthermore, a quantitative correlation model between crack fractal characteristics and structural damage severity is established by combining a U-Net segmentation network with the differential box-counting algorithm. The experimental results demonstrate that under a cyclic loading of 147.6–221.4 kN, the fractal dimension monotonically increases from 1.073 (moderate damage) to 1.189 (failure), with 100% accuracy in damage state identification, closely aligning with the degradation trend of macroscopic mechanical properties. In complex crack scenarios, the model attains a recall rate (Re = 0.882), surpassing U-Net by 13.9%, with significantly enhanced edge reconstruction precision. Compared with the mainstream models, this method effectively alleviates the problem of data annotation dependence through a semi-supervised strategy while maintaining high accuracy. It provides an efficient structural health monitoring solution for engineering practice, which is of great value to promote the application of intelligent detection technology in infrastructure operation and maintenance. Full article

Consumer interest in vegetarian, ethical, and clean-label foods is reviving the use of plant-derived milk coagulants. Cardosins from Cynara cardunculus (“thistle”) are aspartic proteases with strong clotting activity, yet their technological impact in cheese remains under-explored. This study compared a commercial thistle extract (PC) with traditional bovine rennet rich in chymosin (AC) during manufacture and 60-day ripening of Caciotta cheese. Classical compositional assays (ripening index, texture profile, color, solubility) were integrated with scanning electron microscopy, three-dimensional surface reconstruction, and descriptive sensory analysis. AC cheeses displayed slower but sustained proteolysis, yielding a higher and more linear ripening index, softer body, greater solubility, and brighter, more yellow appearance. Imaging revealed a continuous protein matrix with uniformly distributed, larger pores, consistent with a dairy-like sensory profile dominated by milky and umami notes. Conversely, PC cheeses underwent rapid early proteolysis that plateaued, producing firmer, chewier curds with lower solubility and darker color. Micrographs showed a fragmented matrix with smaller, heterogeneous pores; sensory evaluation highlighted vegetal, bitter, and astringent attributes. The data demonstrate that thistle coagulant can successfully replace animal rennet but generates cheeses with distinct structural and sensory fingerprints. The optimization of process parameters is therefore required when targeting specific product styles. Full article

Global warming and anthropogenic climate change are projected to expand the geographic distribution and population abundance of ectothermic species and exacerbate the biological invasion of exotic species. DNA methylation, as a reversible epigenetic modification, could provide a putative link between the phenotypic plasticity of invasive species and environmental temperature variations. We assessed and interpreted the epigenetic mechanisms of invasive and indigenous species’ differential tolerance to thermal stress through the invasive species Bemisia tabaci Mediterranean (MED) and the indigenous species Bemisia tabaci AsiaII3. We examine their thermal tolerance following exposure to heat and cold stress. We found that MED exhibits higher thermal resistance than AsiaII3 under heat stress. The fluorescence-labeled methylation-sensitive amplified polymorphism (F-MSAP) results proved that the increased thermal tolerance in MED is closely related to DNA methylation changes, other than genetic variation. Furthermore, the quantitative real-time polymerase chain reaction (qRT-PCR) and Western blotting analysis of DNA methyltransferases (Dnmts) suggested that increased expression of Dnmt3 regulates the higher thermal tolerance of female MED adults. A mechanism is revealed whereby DNA methylation enhances thermal tolerance in invasive species. Our results show that the Dnmt-mediated regulation mechanism is particularly significant for understanding invasive species’ successful invasion and rapid adaptation under global warming, providing new potential targets for controlling invasive species worldwide. Full article

Tea cultivation faces significant threats from red spider mite (RSM: Oligonychus coffeae) infestations, which reduce yields and economic viability in major tea-producing regions. Current automated detection methods rely on supervised deep learning models requiring extensive labeled data, limiting scalability for smallholder farmers. This article proposes AIM-Net (AI-based Infestation Mapping Network) by evaluating SwAV (Swapping Assignments between Views), a self-supervised learning framework, for classifying RSM infestation severity (Mild, Moderate, Severe) using a geo-referenced, field-acquired dataset of RSM infested tea-leaves, Cam-RSM. The methodology combines SwAV pre-training on unlabeled data with fine-tuning on labeled subsets, employing multi-crop augmentation and online clustering to learn discriminative features without full supervision. Comparative analysis against a fully supervised ResNet-50 baseline utilized 5-fold cross-validation, assessing accuracy, F1-scores, and computational efficiency. Results demonstrate SwAV’s superiority, achieving 98.7% overall accuracy (vs. 92.1% for ResNet-50) and macro-average F1-scores of 98.3% across classes, with a 62% reduction in labeled data requirements. The model showed particular strength in Mild_RSM-class detection (F1-score: 98.5%) and computational efficiency, enabling deployment on edge devices. Statistical validation confirmed significant improvements (p < 0.001) over baseline approaches. These findings establish self-supervised learning as a transformative tool for precision pest management, offering resource-efficient solutions for early infestation detection while maintaining high accuracy. Full article

Aggressiveness in dogs is a complex behavioral trait with implications for animal welfare and public safety. Despite domestication, dogs retain aggressive tendencies shaped by both genetic and environmental factors. This systematic review synthesizes the literature from 2000 to 2024 on the genetic and environmental bases of canine aggression. Using PRISMA 2020 guidelines, 144 articles were retrieved from Scopus and PubMed and screened in two phases, resulting in 33 studies selected for analysis. These were evaluated using a 20-question grid across seven categories, including phenotyping, genetic analysis, population structure, and future directions. The studies support a polygenic model of aggressiveness, with associations reported for genes involved in neurotransmission, hormone signaling, and brain function. However, inconsistencies in phenotyping, small sample sizes, and a limited consideration of environmental factors hinder robust conclusions. Most studies focused on popular companion breeds, while those commonly labeled as aggressive were underrepresented. The findings highlight the relevance of gene–environment interactions but underscore that aggression is often poorly defined and measured across studies. Future research should prioritize standardized phenotyping tools, broader breed inclusion, and the functional validation of genetic findings. These efforts will improve the understanding of dog aggression and inform breeding, behavioral assessment, and public policy. Full article

Background: Statins exhibit pleiotropic anti-inflammatory, antioxidant, and immunomodulatory effects, suggesting their potential in non-cardiovascular conditions. However, evidence supporting their repurposing remains limited, and off-label prescribing policies vary globally. Objective: To systematically review evidence on statin repurposing in oncology and infectious diseases, and to assess Hungarian regulatory practices regarding off-label statin use. Methods: A systematic literature search (PubMed, Web of Science, Scopus, ScienceDirect; 2010–May 2025) was conducted using the terms “drug repositioning” OR “off-label prescription” AND “statin” NOT “cardiovascular,” following PRISMA guidelines. Hungarian off-label usage data from the NNGYK (2008–2025) were also analyzed. Results: Out of 205 publications, 12 met the inclusion criteria—75% were oncology-focused, and 25% focused on infectious diseases. Most were preclinical (58%); only 25% offered strong clinical evidence. Applications included hematologic malignancies, solid tumors, Cryptococcus neoformans, SARS-CoV-2, and dengue virus. Mechanisms involved mevalonate pathway inhibition and modulation of host immune responses. Hungarian data revealed five approved off-label statin uses—three dermatologic and two pediatric metabolic—supported by the literature and requiring post-treatment reporting. Conclusions: While preclinical findings are promising, clinical validation of off-label statin use remains limited. Statins should be continued in cancer patients with cardiovascular indications, but initiation for other purposes should be trial-based. Future directions include biomarker-based personalization, regulatory harmonization, and cost-effectiveness studies. Full article

The lateral flow immunoassay (LFIA) is a widely utilized, rapid diagnostic technique characterized by its short analysis duration, cost efficiency, visual result interpretation, portability and suitability for point-of-care applications. However, conventional LFIAs have limited sensitivity, a challenge that can be overcome by the introduction of gold nanoparticles, which provide enhanced sensitivity and selectivity (compared, for example, to latex beads or carbon nanoparticles) for the detection of target analytes, due to their optical properties, chemical stability and ease of functionalization. In this work, gold nanoparticle-based LFIAs are developed for the detection of aflatoxin B1, and the relative performance of different morphology particles is evaluated. LFIA using gold nano-labels allowed for aflatoxin B1 detection over a range of 0.01 ng/mL–100 ng/mL. Compared to spherical gold nanoparticles and gold nano-flowers, star-shaped gold nanoparticles show increased antibody binding efficiency of 86% due to their greater surface area. Gold nano-stars demonstrated the highest sensitivity, achieving a limit of detection of 0.01ng/mL, surpassing the performance of both spherical gold nanoparticles and gold nano-flowers. The use of star-shaped particles as nano-labels has demonstrated a five-fold improvement in sensitivity, underscoring the potential of integrating diverse nanostructures into LFIA for significantly improving analyte detection. Moreover, the robustness and feasibility of gold nano-stars employed as labels in LFIA was assessed in detecting aflatoxin B1 in a wheat matrix. Improved sensitivity with gold nano-stars holds promise for applications in food safety monitoring, public health diagnostics and rapid point-of-care diagnostics. This work opens the pathway for further development of LFIA utilizing novel nanostructures to achieve unparallel precision in diagnostics and sensing. Full article

Since the initial use of biological ion channels to detect single-stranded genomic base pair differences, label-free and highly sensitive resistive pulse sensing (RPS) with nanopores has made remarkable progress in single-molecule analysis. By monitoring transient ionic current disruptions caused by molecules translocating through a nanopore, this technology offers detailed insights into the structure, charge, and dynamics of the analytes. In this work, the RPS platforms based on biological, solid-state, and other sensing pores, detailing their latest research progress and applications, are reviewed. Their core capability is the high-precision characterization of tiny particles, ions, and nucleotides, which are widely used in biomedicine, clinical diagnosis, and environmental monitoring. However, current RPS methods involve bottlenecks, including limited sensitivity (weak signals from sub-nanometer targets with low SNR), complex sample interference (high false positives from ionic strength, etc.), and field consistency (solid-state channel drift, short-lived bio-pores failing POCT needs). To overcome this, bio-solid-state fusion channels, in-well reactors, deep learning models, and transfer learning provide various options. Evolving into an intelligent sensing ecosystem, RPS is expected to become a universal platform linking basic research, precision medicine, and on-site rapid detection. Full article

This study evaluates the efficacy of commercial clean-label additives, specifically fermentates, in inhibiting mold growth in vitro and extending the shelf life of preservative-free bread. The mold growth on selected bread was modeled using the time-to-growth approach. The pH, a_w, and moisture content of fresh bread were determined. In addition, selected fermentates were characterized physicochemically. Fermentates, defined as liquid or powdered preparations containing microorganisms, their metabolites, and culture supernatants, were tested at varying concentrations (1% to 12%) to assess their antimicrobial performance and impact on bread quality parameters, including moisture content, water activity, and pH. The results showed significant differences in fermentate efficacy, with Product A as the best mold growth inhibitor in vitro and a clear dose-dependent response. For Penicillium corylophilum, inhibition increased from 51.90% at 1% to 62.60% at 4%, while P. chrysogenum had an inhibition ranging from 32.26% to 34.49%. Product F exhibited moderate activity on both molds at 4%, inhibiting between 28.48% and 46.27%. The two molds exhibited differing sensitivities to the fermentates, with P. corylophilum consistently more susceptible to inhibition. Product A displayed a low pH (2.61) and high levels of lactic acid (1053.6 mmol/L) and acetic acid (1061.3 mmol/L). Product F presented a similar pH but lower levels of lactic and acetic acid. A time-to-growth model, validated by significant coefficients (p < 0.05) and high predictive accuracy (R² > 0.95), was employed to predict the appearance of mold on bread loaves. The model revealed that higher concentrations of fermentates A and F delayed mold growth, with fermentate A demonstrating superior efficacy. At 2% concentration, fermentate A delayed mold growth for 8 days, compared to 6 days for fermentate F. At 8% concentration, fermentate A prevented mold growth for over 25 days, significantly outperforming the control (4 days). Additionally, fermentates influenced bread quality parameters, with fermentate A improving crust moisture retention and reducing water activity at higher concentrations. These findings highlight the potential of fermentates as sustainable, consumer-friendly alternatives to synthetic preservatives, offering a viable solution to the challenge of bread spoilage while maintaining product quality. Full article

Background/Objectives: Cannabidiol (CBD) is a non-intoxicating cannabinoid touted for a variety of medical benefits, including alleviation of anxiety. While legalization of hemp-derived products in the United States (containing ≤0.3% delta-9-tetrahydrocannabinol [d9-THC] by weight) has led to a rapid increase in the commercialization of hemp-derived CBD products, most therapeutic claims have not been substantiated using clinical trials. This trial aimed to assess the impact of 6 weeks of treatment with a proprietary hemp-derived, full-spectrum, high-CBD sublingual solution similar to those available in the marketplace in patients with anxiety. Methods: An open-label pilot clinical trial (NCT04286594) was conducted in 12 patients with at least moderate levels of anxiety. Patients self-administered a hemp-derived, high-CBD sublingual solution twice daily during the 6-week trial (target daily dose: 30 mg/day CBD). Clinical change over time relative to baseline was assessed for anxiety, mood, sleep, and quality of life, as well as changes in cognitive performance on measures of executive function and memory. Safety and tolerability of the study product were also evaluated. Results: Patients reported significant reductions in anxiety symptoms over time. Concurrent improvements in mood, sleep, and relevant quality of life domains were also observed, along with stable or improved performance on all neurocognitive measures. Few side effects were reported, and no serious adverse events occurred. Conclusions: These pilot findings provide initial support for the efficacy and tolerability of the hemp-derived, high-CBD product in patients with moderate-to-severe levels of anxiety. Double-blind, placebo-controlled studies are indicated to obtain robust data regarding efficacy and tolerability of these types of products for anxiety. Full article

Compared to traditional real-valued neural networks, which process only amplitude information, complex-valued neural networks handle both amplitude and phase information, leading to superior performance in polarimetric synthetic aperture radar (PolSAR) image classification tasks. This paper proposes a multi-scale feature extraction (MSFE) method based on a 3D complex-valued network to improve classification accuracy by fully leveraging multi-scale features, including phase information. We first designed a complex-valued three-dimensional network framework combining complex-valued 3D convolution (CV-3DConv) with complex-valued squeeze-and-excitation (CV-SE) modules. This framework is capable of simultaneously capturing spatial and polarimetric features, including both amplitude and phase information, from PolSAR images. Furthermore, to address robustness degradation from limited labeled samples, we introduced a multi-scale learning strategy that jointly models global and local features. Specifically, global features extract overall semantic information, while local features help the network capture region-specific semantics. This strategy enhances information utilization by integrating multi-scale receptive fields, complementing feature advantages. Extensive experiments on four benchmark datasets demonstrated that the proposed method outperforms various comparison methods, maintaining high classification accuracy across different sampling rates, thus validating its effectiveness and robustness. Full article

Work-related musculoskeletal disorders (WMSDs) are a leading concern in industrial ergonomics, often stemming from sustained non-neutral postures and repetitive tasks. This paper presents a vision-based framework for real-time, frame-level ergonomic risk classification using a lightweight Vision Transformer (ViT). The proposed system operates directly on raw RGB images without requiring skeleton reconstruction, joint angle estimation, or image segmentation. A single ViT model simultaneously classifies eight anatomical regions, enabling efficient multi-label posture assessment. Training is supervised using a multimodal dataset acquired from synchronized RGB video and full-body inertial motion capture, with ergonomic risk labels derived from RULA scores computed on joint kinematics. The system is validated on realistic, simulated industrial tasks that include common challenges such as occlusion and posture variability. Experimental results show that the ViT model achieves state-of-the-art performance, with F1-scores exceeding 0.99 and AUC values above 0.996 across all regions. Compared to previous CNN-based system, the proposed model improves classification accuracy and generalizability while reducing complexity and enabling real-time inference on edge devices. These findings demonstrate the model’s potential for unobtrusive, scalable ergonomic risk monitoring in real-world manufacturing environments. Full article