With the increase of semiconductor integration density,in order to cope with the increase of wafer defect complexity and types,especially the low recognition accuracy of overlapping mixed defects and unknown wafer def...With the increase of semiconductor integration density,in order to cope with the increase of wafer defect complexity and types,especially the low recognition accuracy of overlapping mixed defects and unknown wafer defects,this study proposes a lightweight model for wafer defect detection called LightWMNet.First,using a hierarchical attention Encoder-Decoder architecture,the features of wafer defect pattern(WDP)are channel recalibrated to generate high-resolution fine-grained features and low-resolution coarse-grained features.Secondly,the backbone network incorporates two novel attention modules—feedforward spatial attention(FFSa)and feedforward channel attention(FFCa)—to amplify responses in critical defect regions and suppress noise from stochastic discrete pixels.These mechanisms synergistically enhance feature discriminability without introducing significant parametric overhead.Finally,the Dice loss function and the cross entropy loss function are combined to jointly evaluate the segmentation and classification accuracy of the model.Experimental results on the public mixed wafer defect dataset MixedWM38 show that the pixel accuracy(PA),intersection over union(IoU)and Dice coefficient of the proposed network reach 98.26%,94.83%and 97.22%,respectively.Without significantly increasing the computational complexity and size of the model,compared with the existing state-of-the-art(SOTA)model,the classification accuracy of lightWMNet in single defect,three mixed defects and four mixed defects is improved by 0.5%,0.25%and 0.89%respectively.Furthermore,we used transfer learning for the first time to evaluate the model's generalisation ability for unseen defect categories.The results showed that LightWMNet still has a certain recognition ability even in untrained wafer defects.展开更多
Background:Accurate classification of brain tumors from Magnetic Resonance Imaging(MRI)is essential for clinical decision-making but remains challenging due to tumor heterogeneity.Existing approaches often focus solel...Background:Accurate classification of brain tumors from Magnetic Resonance Imaging(MRI)is essential for clinical decision-making but remains challenging due to tumor heterogeneity.Existing approaches often focus solely on classification or treat segmentation and classification as separate tasks,limiting overall performance and interpretability.Methods:This study proposes an end-to-end automated framework that integrates optimized tumor localization with multiclass classification.An optimized segmentation model is first employed to generate tumor masks,which are then overlaid on MRI scans to produce attention-enhanced inputs.These inputs are subsequently used to train a convolutional neural network(CNN)classifier.Experiments were conducted on a public dataset comprising 4,237 MRI scans across four categories:normal,glioma,meningioma,and pituitary tumors.Results:Three widely used segmentation models were systematically evaluated,with an optimized U-Net achieving the best performance(accuracy=0.9939,Dice=0.8893).Segmentation-guided classification consistently improved performance across six CNN architectures,with the most notable gains observed in heterogeneous tumor types such as glioma and meningioma.Among the classifiers,EfficientNet-V2 achieved the highest performance,with an accuracy of 0.9835,precision of 0.9858,recall of 0.9804,and F1-score of 0.9828.The framework was further validated on an independent external dataset,demonstrating consistent performance and robustness across diverse MRI sources.Conclusion:The proposed framework demonstrates strong potential for multiclass brain tumor classification by effectively combining segmentation and classification.This segmentation-driven approach not only enhances predictive accuracy but also improves interpretability,making it more suitable for clinical applications.展开更多
In recent decades,the proliferation of email communication has markedly escalated,resulting in a concomitant surge in spam emails that congest networks and presenting security risks.This study introduces an innovative...In recent decades,the proliferation of email communication has markedly escalated,resulting in a concomitant surge in spam emails that congest networks and presenting security risks.This study introduces an innovative spam detection method utilizing the Horse Herd Optimization Algorithm(HHOA),designed for binary classification within multi⁃objective framework.The method proficiently identifies essential features,minimizing redundancy and improving classification precision.The suggested HHOA attained an impressive accuracy of 97.21%on the Kaggle email dataset,with precision of 94.30%,recall of 90.50%,and F1⁃score of 92.80%.Compared to conventional techniques,such as Support Vector Machine(93.89%accuracy),Random Forest(96.14%accuracy),and K⁃Nearest Neighbours(92.08%accuracy),HHOA exhibited enhanced performance with reduced computing complexity.The suggested method demonstrated enhanced feature selection efficiency,decreasing the number of selected features while maintaining high classification accuracy.The results underscore the efficacy of HHOA in spam identification and indicate its potential for further applications in practical email filtering systems.展开更多
Visual diagnosis of skin cancer is challenging due to subtle inter-class similarities,variations in skin texture,the presence of hair,and inconsistent illumination.Deep learning models have shown promise in assisting ...Visual diagnosis of skin cancer is challenging due to subtle inter-class similarities,variations in skin texture,the presence of hair,and inconsistent illumination.Deep learning models have shown promise in assisting early detection,yet their performance is often limited by the severe class imbalance present in dermoscopic datasets.This paper proposes CANNSkin,a skin cancer classification framework that integrates a convolutional autoencoder with latent-space oversampling to address this imbalance.The autoencoder is trained to reconstruct lesion images,and its latent embeddings are used as features for classification.To enhance minority-class representation,the Synthetic Minority Oversampling Technique(SMOTE)is applied directly to the latent vectors before classifier training.The encoder and classifier are first trained independently and later fine-tuned end-to-end.On the HAM10000 dataset,CANNSkin achieves an accuracy of 93.01%,a macro-F1 of 88.54%,and an ROC–AUC of 98.44%,demonstrating strong robustness across ten test subsets.Evaluation on the more complex ISIC 2019 dataset further confirms the model’s effectiveness,where CANNSkin achieves 94.27%accuracy,93.95%precision,94.09%recall,and 99.02%F1-score,supported by high reconstruction fidelity(PSNR 35.03 dB,SSIM 0.86).These results demonstrate the effectiveness of our proposed latent-space balancing and fine-tuned representation learning as a new benchmark method for robust and accurate skin cancer classification across heterogeneous datasets.展开更多
Artificial Intelligence(AI)in healthcare enables predicting diabetes using data-driven methods instead of the traditional ways of screening the disease,which include hemoglobin A1c(HbA1c),oral glucose tolerance test(O...Artificial Intelligence(AI)in healthcare enables predicting diabetes using data-driven methods instead of the traditional ways of screening the disease,which include hemoglobin A1c(HbA1c),oral glucose tolerance test(OGTT),and fasting plasma glucose(FPG)screening techniques,which are invasive and limited in scale.Machine learning(ML)and deep neural network(DNN)models that use large datasets to learn the complex,nonlinear feature interactions,but the conventional ML algorithms are data sensitive and often show unstable predictive accuracy.Conversely,DNN models are more robust,though the ability to reach a high accuracy rate consistently on heterogeneous datasets is still an open challenge.For predicting diabetes,this work proposed a hybrid DNN approach by integrating a bidirectional long short-term memory(BiLSTM)network with a bidirectional gated recurrent unit(BiGRU).A robust DL model,developed by combining various datasets with weighted coefficients,dense operations in the connection of deep layers,and the output aggregation using batch normalization and dropout functions to avoid overfitting.The goal of this hybrid model is better generalization and consistency among various datasets,which facilitates the effective management and early intervention.The proposed DNN model exhibits an excellent predictive performance as compared to the state-of-the-art and baseline ML and DNN models for diabetes prediction tasks.The robust performance indicates the possible usefulness of DL-based models in the development of disease prediction in healthcare and other areas that demand high-quality analytics.展开更多
As large-scale astronomical surveys,such as the Sloan Digital Sky Survey(SDSS)and the Large Sky Area Multi-Object Fiber Spectroscopic Telescope(LAMOST),generate increasingly complex datasets,clustering algorithms have...As large-scale astronomical surveys,such as the Sloan Digital Sky Survey(SDSS)and the Large Sky Area Multi-Object Fiber Spectroscopic Telescope(LAMOST),generate increasingly complex datasets,clustering algorithms have become vital for identifying patterns and classifying celestial objects.This paper systematically investigates the application of five main categories of clustering techniques-partition-based,density-based,model-based,hierarchical,and“others”-across a range of astronomical research over the past decade.This review focuses on the six key application areas of stellar classification,galaxy structure analysis,detection of galactic and interstellar features,highenergy astrophysics,exoplanet studies,and anomaly detection.This paper provides an in-depth analysis of the performance and results of each method,considering their respective suitabilities for different data types.Additionally,it presents clustering algorithm selection strategies based on the characteristics of the spectroscopic data being analyzed.We highlight challenges such as handling large datasets,the need for more efficient computational tools,and the lack of labeled data.We also underscore the potential of unsupervised and semi-supervised clustering approaches to overcome these challenges,offering insight into their practical applications,performance,and results in astronomical research.展开更多
The Chamidae are widely distributed in the tropical to temperate seas,with>70 known species.Currently,their classification relies mainly on traditional morphological methods and identification primarily on small fr...The Chamidae are widely distributed in the tropical to temperate seas,with>70 known species.Currently,their classification relies mainly on traditional morphological methods and identification primarily on small fragment genes,such as COI.The intrafamily phylogenetic relationships are ambiguous,lacking support from reliable molecular data.In this study,the mitochondrial genomes of eight species of Chamidae were sequenced for the first time and then annotated.Their structures and compositional characteristics were analyzed.The mitochondrial gene order in this family differed significantly.Concurrently,the evolutionary position and phylogenetic relationship among Chamidae species were explored,and the Veneroida phylogenetic tree was recreated.Maximum Likelihood and Bayesian Inference analyses supported the monophyly of Chamidae.Additionally,the divergence time within Chamidae was estimated based on mitochondrial DNA sequences,with the most ancient divergence estimated to occur during the early Cretaceous period,128 MYA.This phylogeny is significant for understanding the diversity and taxonomic status of Chamidae.展开更多
The genus Clematis(Ranunculaceae)comprises over 300 species with remarkable morphological and ecological diversity worldwide.Despite its horticultural,medicinal,and ecological importance,a well-resolved phylogeny and ...The genus Clematis(Ranunculaceae)comprises over 300 species with remarkable morphological and ecological diversity worldwide.Despite its horticultural,medicinal,and ecological importance,a well-resolved phylogeny and coherent infrageneric classification are still lacking.Here,we reconstruct a robust phylogeny for Clematis using a phylogenomic approach and revise its infrageneric taxonomy.We incorporated 198 samples representing 151 species,two subspecies,and 12 varieties,covering all subgenera and most sections worldwide,obtained from both fresh and herbarium material.Nuclear single nucleotide polymorphisms(SNPs)and complete plastid genomes were assembled for phylogenetic analyses.We also prepared a nuclear ribosomal ITS(nrITS)dataset comprising 171 species,two subspecies,and 12 varieties(217 samples)to include as many species as possible for phylogenetic inference.Phylogenies based on plastid genomes and nrITS exhibited limited resolution and modest support,highlighting challenges in resolving certain relationships.Nuclear SNP analyses yielded a robust phylogenetic tree with 22 well-supported clades corresponding to 22 sections,with most previously recognized subgenera and sections not recovered as monophyletic.Ancestral state reconstruction of 12 key morphological characters revealed multiple independent origins of character states.This study presents the first comprehensive sectional classification for Clematis based on robust phylogenomic evidence,redefines morphological characteristics for each section,and resolves long-standing taxonomic ambiguities.Our results establish a framework for future studies on the evolution,ecology,and horticultural potential of this globally significant genus.展开更多
Classifying job offers into occupational categories is a fundamental task in human resource information systems,as it improves and streamlines indexing,search,and matching between openings and job seekers.Comprehensiv...Classifying job offers into occupational categories is a fundamental task in human resource information systems,as it improves and streamlines indexing,search,and matching between openings and job seekers.Comprehensive occupational databases such as O∗NET or ESCO provide detailed taxonomies of interrelated positions that can be leveraged to align the textual content of postings with occupational categories,thereby facilitating standardization,cross-system interoperability,and access to metadata for each occupation(e.g.,tasks,knowledge,skills,and abilities).In this work,we explore the effectiveness of fine-tuning existing language models(LMs)to classify job offers with occupational descriptors from O∗NET.This enables a more precise assessment of candidate suitability by identifying the specific knowledge and skills required for each position,and helps automate recruitment processes by mitigating human bias and subjectivity in candidate selection.We evaluate three representative BERT-like models:BERT,RoBERTa,and DeBERTa.BERT serves as the baseline encoder-only architecture;RoBERTa incorporates advances in pretraining objectives and data scale;and DeBERTa introduces architectural improvements through disentangled attention mechanisms.The best performance was achieved with the DeBERTa model,although the other models also produced strong results,and no statistically significant differences were observed acrossmodels.We also find that these models typically reach optimal performance after only a few training epochs,and that training with smaller,balanced datasets is effective.Consequently,comparable results can be obtained with models that require fewer computational resources and less training time,facilitating deployment and practical use.展开更多
With the increasing complexity of malware attack techniques,traditional detection methods face significant challenges,such as privacy preservation,data heterogeneity,and lacking category information.To address these i...With the increasing complexity of malware attack techniques,traditional detection methods face significant challenges,such as privacy preservation,data heterogeneity,and lacking category information.To address these issues,we propose Federated Dynamic Prototype Learning(FedDPL)for malware classification by integrating Federated Learning with a specifically designed K-means.Under the Federated Learning framework,model training occurs locally without data sharing,effectively protecting user data privacy and preventing the leakage of sensitive information.Furthermore,to tackle the challenges of data heterogeneity and the lack of category information,FedDPL introduces a dynamic prototype learning mechanism,which adaptively adjusts the clustering prototypes in terms of position and number.Thus,the dependency on predefined category numbers in typical K-means and its variants can be significantly reduced,resulting in improved clustering performance.Theoretically,it provides a more accurate detection of malicious behavior.Experimental results confirm that FedDPL excels in handling malware classification tasks,demonstrating superior accuracy,robustness,and privacy protection.展开更多
Arrhythmias are a frequently occurring phenomenon in clinical practice,but how to accurately dis-tinguish subtle rhythm abnormalities remains an ongoing difficulty faced by the entire research community when conductin...Arrhythmias are a frequently occurring phenomenon in clinical practice,but how to accurately dis-tinguish subtle rhythm abnormalities remains an ongoing difficulty faced by the entire research community when conducting ECG-based studies.From a review of existing studies,two main factors appear to contribute to this problem:the uneven distribution of arrhythmia classes and the limited expressiveness of features learned by current models.To overcome these limitations,this study proposes a dual-path multimodal framework,termed DM-EHC(Dual-Path Multimodal ECG Heartbeat Classifier),for ECG-based heartbeat classification.The proposed framework links 1D ECG temporal features with 2D time–frequency features.By setting up the dual paths described above,the model can process more dimensions of feature information.The MIT-BIH arrhythmia database was selected as the baseline dataset for the experiments.Experimental results show that the proposed method outperforms single modalities and performs better for certain specific types of arrhythmias.The model achieved mean precision,recall,and F1 score of 95.14%,92.26%,and 93.65%,respectively.These results indicate that the framework is robust and has potential value in automated arrhythmia classification.展开更多
The rapid digitalization of the energy sector has led to the deployment of large-scale smart metering systems that generate high-frequency time series data,creating new opportunities and challenges for energy anomaly ...The rapid digitalization of the energy sector has led to the deployment of large-scale smart metering systems that generate high-frequency time series data,creating new opportunities and challenges for energy anomaly detection.Accurate identification of anomalous patterns in building energy consumption is essential for optimizing operations,improving energy efficiency,and supporting grid reliability.This study investigates advanced feature engineering and machine learning modeling techniques for large-scale time series anomaly detection in building energy systems.Expanding upon previous benchmark frameworks,we introduce additional features such as oil price indices and solar cycle indicators,including sunset and sunrise times,to enhance the contextual understanding of consumption patterns.Our comparative modeling approach encompasses an extensive suite of algorithms,including KNeighborsUnif,KNeighborsDist,LightGBMXT,LightGBM,RandomForestMSE,CatBoost,ExtraTreesMSE,NeuralNetFastAI,XGBoost,NeuralNetTorch,and LightGBMLarge.Data preprocessing includes rigorous handling of missing values and normalization,while feature engineering focuses on temporal,environmental,and value-change attributes.The models are evaluated on a comprehensive dataset of smart meter readings,with performance assessed using metrics such as the Area Under the Receiver Operating Characteristic Curve(AUC-ROC).The results demonstrate that the integration of diverse exogenous variables and a hybrid ensemble of traditional tree-based and neural network models can significantly improve anomaly detection performance.This work provides new insights into the design of robust,scalable,and generalizable frameworks for energy anomaly detection in complex,real-world settings.展开更多
Spam emails remain one of the most persistent threats to digital communication,necessitating effective detection solutions that safeguard both individuals and organisations.We propose a spam email classification frame...Spam emails remain one of the most persistent threats to digital communication,necessitating effective detection solutions that safeguard both individuals and organisations.We propose a spam email classification frame-work that uses Bidirectional Encoder Representations from Transformers(BERT)for contextual feature extraction and a multiple-window Convolutional Neural Network(CNN)for classification.To identify semantic nuances in email content,BERT embeddings are used,and CNN filters extract discriminative n-gram patterns at various levels of detail,enabling accurate spam identification.The proposed model outperformed Word2Vec-based baselines on a sample of 5728 labelled emails,achieving an accuracy of 98.69%,AUC of 0.9981,F1 Score of 0.9724,and MCC of 0.9639.With a medium kernel size of(6,9)and compact multi-window CNN architectures,it improves performance.Cross-validation illustrates stability and generalization across folds.By balancing high recall with minimal false positives,our method provides a reliable and scalable solution for current spam detection in advanced deep learning.By combining contextual embedding and a neural architecture,this study develops a security analysis method.展开更多
Malware poses a significant threat to the Internet of Things(IoT).It enables unauthorized access to devices in the IoT environment.The lack of unique architectural standards causes challenges in developing robust malw...Malware poses a significant threat to the Internet of Things(IoT).It enables unauthorized access to devices in the IoT environment.The lack of unique architectural standards causes challenges in developing robust malware detection(MD)models.The existing models demand substantial computational resources.This study intends to build a lightweight MD model to detect anomalies in IoT networks.The authors develop a transformation technique,converting the malware binaries into images.MobileNet V2 is fine-tuned using improved grey wolf optimization(IGWO)to extract crucial features of malicious and benign samples.The ResNeXt model is combined with the Linformer’s attention mechanism to identify Malware features.A fully connected layer is integrated with gradientweighted class activation mapping(Grad-CAM)in order to facilitate an interpretable classification model.The proposed model is evaluated using the IoT malware and the IoT-23 datasets.The model performs well on the two datasets with an accuracy of 98.94%,precision of 98.46%,recall of 98.11%,and F1-score of 98.28%on the IoT malware dataset,and an accuracy of 98.23%,precision of 96.80%,recall of 96.64%,and F1-score of 96.71%on the IoT-23 dataset,respectively.The findings indicate that the model has a high standard of classification.The lightweight architecture enables efficient deployment with an inference time of 1.42 s.Inference time has no direct impact on accuracy,precision,recall,or F1-score.However,the inference speed would warrant timely detection in latency-sensitive IoT applications.By achieving a remarkable result,the proposed study offers a comprehensive solution:a scalable,interpretable,and computationally efficient MD model for the evolving IoT landscape.展开更多
Human Activity Recognition(HAR)is a novel area for computer vision.It has a great impact on healthcare,smart environments,and surveillance while is able to automatically detect human behavior.It plays a vital role in ...Human Activity Recognition(HAR)is a novel area for computer vision.It has a great impact on healthcare,smart environments,and surveillance while is able to automatically detect human behavior.It plays a vital role in many applications,such as smart home,healthcare,human computer interaction,sports analysis,and especially,intelligent surveillance.In this paper,we propose a robust and efficient HAR system by leveraging deep learning paradigms,including pre-trained models,CNN architectures,and their average-weighted fusion.However,due to the diversity of human actions and various environmental influences,as well as a lack of data and resources,achieving high recognition accuracy remain elusive.In this work,a weighted average ensemble technique is employed to fuse three deep learning models:EfficientNet,ResNet50,and a custom CNN.The results of this study indicate that using a weighted average ensemble strategy for developing more effective HAR models may be a promising idea for detection and classification of human activities.Experiments by using the benchmark dataset proved that the proposed weighted ensemble approach outperformed existing approaches in terms of accuracy and other key performance measures.The combined average-weighted ensemble of pre-trained and CNN models obtained an accuracy of 98%,compared to 97%,96%,and 95%for the customized CNN,EfficientNet,and ResNet50 models,respectively.展开更多
In this study,we explore a spherically symmetric charged black hole(BH)with a negative cosmological constant under the influence of a Kalb-Ramond field background.We compute the photon sphere and shadow radii,validati...In this study,we explore a spherically symmetric charged black hole(BH)with a negative cosmological constant under the influence of a Kalb-Ramond field background.We compute the photon sphere and shadow radii,validating our findings using observational data from the Event Horizon Telescope,with a particular emphasis on the shadow images of Sagittarius A^(*).Furthermore,we investigate the greybody factors,emission rate,and partial absorption cross section.It is shown that the Lorentz-violating parameter l has an important effect on the absorption cross section.Our analysis also includes an examination of the topological charge,temperature-dependent topology,and generalized free energy.In particular,we regard the AdS charged BH with an antisymmetric tensor background as a topological defect in the thermodynamic space,then the system has the same topological classification to the charged Reissner-Nordström-AdS BH.展开更多
Designing appropriate loss functions is critical to the success of supervised learning models.However,most conventional losses are fixed and manually designed,making them suboptimal for diverse and dynamic learning sc...Designing appropriate loss functions is critical to the success of supervised learning models.However,most conventional losses are fixed and manually designed,making them suboptimal for diverse and dynamic learning scenarios.In this work,we propose an Adaptive Meta-Loss Network(Adaptive-MLN)that learns to generate taskagnostic loss functions tailored to evolving classification problems.Unlike traditional methods that rely on static objectives,Adaptive-MLN treats the loss function itself as a trainable component,parameterized by a shallow neural network.To enable flexible,gradient-free optimization,we introduce a hybrid evolutionary approach that combines GeneticAlgorithms(GA)for global exploration and Evolution Strategies(ES)for local refinement.This co-evolutionary process dynamically adjusts the loss landscape,improvingmodel generalization without relying on analytic gradients or handcrafted heuristics.Experimental evaluations on synthetic tasks and the CIFAR-10 andMNIST datasets demonstrate that our approach consistently outperforms standard losses such as Cross-Entropy and Mean Squared Error in terms of accuracy,convergence,and adaptability.展开更多
The fine-scale characterization of vegetation surface information serves as a fundamental basis for studying the spatial distribution of resources and the dynamic patterns of environmental responses.Accurately extract...The fine-scale characterization of vegetation surface information serves as a fundamental basis for studying the spatial distribution of resources and the dynamic patterns of environmental responses.Accurately extracting the distributions of different crop species is of critical importance for improving agricultural production efficiency and ensuring food security.Traditional fine-scale vegetation extraction methods often face significant challenges due to the presence of spectrally similar features and the substantial influence of background interference,which limit their applicability across large areas.As a key phenological stage of angiosperms,flowering is characterized by distinctive flowering times,floral morphology,and canopy spectral signatures,so it is an effective pathway for fine-scale vegetation extraction using remote sensing.Using rapeseed as an example,this study developed a spectral index model for precise flowering vegetation extraction(FI-R)based on Landsat OLI imagery.The model integrates a yellowness index(Blue,Green)and a peak index(Red,Nir and SWIR1)while leveraging the NDVI to mitigate background interference from spectrally similar objects.This approach successfully enables the rapid and accurate large-scale mapping of flowering vegetation under complex background conditions.The proposed method was tested in five rapeseed cultivation regions worldwide with diverse backgrounds.Validation datasets were generated using GF imagery and the U.S.CDL dataset.The FI-R model demonstrated superior capability in distinguishing flowering rapeseed from other vegetation,and achieved overall accuracies exceeding 94%in all study areas.Furthermore,FI-R is compatible with other multispectral sensors that have similar band configurations,so it is applicable to rapeseed extraction in broader contexts.The method also shows strong potential for the fine-scale extraction of other types of flowering angiosperm vegetation.展开更多
With the recent increase in data volume and diversity,traditional text representation techniques are struggling to capture context,particularly in environments with sparse data.To address these challenges,this study p...With the recent increase in data volume and diversity,traditional text representation techniques are struggling to capture context,particularly in environments with sparse data.To address these challenges,this study proposes a new model,the Masked Joint Representation Model(MJRM).MJRM approximates the original hypothesis by leveraging multiple elements in a limited context.It dynamically adapts to changes in characteristics based on data distribution through three main components.First,masking-based representation learning,termed selective dynamic masking,integrates topic modeling and sentiment clustering to generate and train multiple instances across different data subsets,whose predictions are then aggregated with optimized weights.This design alleviates sparsity,suppresses noise,and preserves contextual structures.Second,regularization-based improvements are applied.Third,techniques for addressing sparse data are used to perform final inference.As a result,MJRM improves performance by up to 4%compared to existing AI techniques.In our experiments,we analyzed the contribution of each factor,demonstrating that masking,dynamic learning,and aggregating multiple instances complement each other to improve performance.This demonstrates that a masking-based multi-learning strategy is effective for context-aware sparse text classification,and can be useful even in challenging situations such as data shortage or data distribution variations.We expect that the approach can be extended to diverse fields such as sentiment analysis,spam filtering,and domain-specific document classification.展开更多
Understanding the spatial distributions and corresponding variation mechanisms of key soil nutrients in fragile karst ecosystems can assist in promoting sustainable development.However,due to the implementation of eco...Understanding the spatial distributions and corresponding variation mechanisms of key soil nutrients in fragile karst ecosystems can assist in promoting sustainable development.However,due to the implementation of ecological restoration initiatives such as land-use conversions,novel changes in the spatial characteristics of soil nutrients remain unknown.To address this gap,we explored nutrient variations and the drivers of the variation in the 0–15 cm topsoil layer using a regional-scale sampling method in a typical karst area in northwest Guangxi Zhuang Autonomous Region,Southwest China.Descriptive statistics,geostatistics,and spatial analysis were used to assess the soil nutrient variability.The results indicated that soil organic carbon(SOC),total nitrogen(TN),total phosphorus(TP),and total potassium(TK)concentrations showed moderate variations,with coefficients of variance being 0.60,0.60,0.71,and 0.72,respectively.Moreover,they demonstrated positive spatial autocorrelations,with global Moran's indices being 0.68,0.77,0.64,and 0.68,respectively.However,local Moran's index values were low,indicating large spatial variations in soil nutrients.The best-fitting semi-variogram models for SOC,TN,TP,and TK concentrations were spherical,Gaussian,exponential,and exponential,respectively.According to the classification criteria of the Second National Soil Census in China,SOC and TN concentrations were relatively sufficient,with the proportions of rich and very rich levels being up to 90.9 and 96.0%,respectively.TP concentration was in the mediumdeficient level,with the areas of medium and deficient levels accounting for 33.7 and 30.1%of the total,respectively.TK concentration was deficient,with the cumulative area of extremely deficient,very deficient,and deficient levels accounting for 87.6%of the total area.Consequently,the terrestrial ecosystems in the study area were more vulnerable to soil P and K than soil N deficiencies.Furthermore,variance partitioning analysis of the influencing factors showed that,except for the interactions,the single effect of other soil properties accounted more for soil nutrient variations than spatial and environmental variables.These results will aid in the future management of terrestrial ecosystems.展开更多
基金supported by the National Natural Science Foundation of China under Grant 61573183.
文摘With the increase of semiconductor integration density,in order to cope with the increase of wafer defect complexity and types,especially the low recognition accuracy of overlapping mixed defects and unknown wafer defects,this study proposes a lightweight model for wafer defect detection called LightWMNet.First,using a hierarchical attention Encoder-Decoder architecture,the features of wafer defect pattern(WDP)are channel recalibrated to generate high-resolution fine-grained features and low-resolution coarse-grained features.Secondly,the backbone network incorporates two novel attention modules—feedforward spatial attention(FFSa)and feedforward channel attention(FFCa)—to amplify responses in critical defect regions and suppress noise from stochastic discrete pixels.These mechanisms synergistically enhance feature discriminability without introducing significant parametric overhead.Finally,the Dice loss function and the cross entropy loss function are combined to jointly evaluate the segmentation and classification accuracy of the model.Experimental results on the public mixed wafer defect dataset MixedWM38 show that the pixel accuracy(PA),intersection over union(IoU)and Dice coefficient of the proposed network reach 98.26%,94.83%and 97.22%,respectively.Without significantly increasing the computational complexity and size of the model,compared with the existing state-of-the-art(SOTA)model,the classification accuracy of lightWMNet in single defect,three mixed defects and four mixed defects is improved by 0.5%,0.25%and 0.89%respectively.Furthermore,we used transfer learning for the first time to evaluate the model's generalisation ability for unseen defect categories.The results showed that LightWMNet still has a certain recognition ability even in untrained wafer defects.
文摘Background:Accurate classification of brain tumors from Magnetic Resonance Imaging(MRI)is essential for clinical decision-making but remains challenging due to tumor heterogeneity.Existing approaches often focus solely on classification or treat segmentation and classification as separate tasks,limiting overall performance and interpretability.Methods:This study proposes an end-to-end automated framework that integrates optimized tumor localization with multiclass classification.An optimized segmentation model is first employed to generate tumor masks,which are then overlaid on MRI scans to produce attention-enhanced inputs.These inputs are subsequently used to train a convolutional neural network(CNN)classifier.Experiments were conducted on a public dataset comprising 4,237 MRI scans across four categories:normal,glioma,meningioma,and pituitary tumors.Results:Three widely used segmentation models were systematically evaluated,with an optimized U-Net achieving the best performance(accuracy=0.9939,Dice=0.8893).Segmentation-guided classification consistently improved performance across six CNN architectures,with the most notable gains observed in heterogeneous tumor types such as glioma and meningioma.Among the classifiers,EfficientNet-V2 achieved the highest performance,with an accuracy of 0.9835,precision of 0.9858,recall of 0.9804,and F1-score of 0.9828.The framework was further validated on an independent external dataset,demonstrating consistent performance and robustness across diverse MRI sources.Conclusion:The proposed framework demonstrates strong potential for multiclass brain tumor classification by effectively combining segmentation and classification.This segmentation-driven approach not only enhances predictive accuracy but also improves interpretability,making it more suitable for clinical applications.
文摘In recent decades,the proliferation of email communication has markedly escalated,resulting in a concomitant surge in spam emails that congest networks and presenting security risks.This study introduces an innovative spam detection method utilizing the Horse Herd Optimization Algorithm(HHOA),designed for binary classification within multi⁃objective framework.The method proficiently identifies essential features,minimizing redundancy and improving classification precision.The suggested HHOA attained an impressive accuracy of 97.21%on the Kaggle email dataset,with precision of 94.30%,recall of 90.50%,and F1⁃score of 92.80%.Compared to conventional techniques,such as Support Vector Machine(93.89%accuracy),Random Forest(96.14%accuracy),and K⁃Nearest Neighbours(92.08%accuracy),HHOA exhibited enhanced performance with reduced computing complexity.The suggested method demonstrated enhanced feature selection efficiency,decreasing the number of selected features while maintaining high classification accuracy.The results underscore the efficacy of HHOA in spam identification and indicate its potential for further applications in practical email filtering systems.
基金supported and funded by the Deanship of Scientific Research at Imam Mohammad Ibn Saud Islamic University(IMSIU)(grant number IMSIU-DDRSP2601).
文摘Visual diagnosis of skin cancer is challenging due to subtle inter-class similarities,variations in skin texture,the presence of hair,and inconsistent illumination.Deep learning models have shown promise in assisting early detection,yet their performance is often limited by the severe class imbalance present in dermoscopic datasets.This paper proposes CANNSkin,a skin cancer classification framework that integrates a convolutional autoencoder with latent-space oversampling to address this imbalance.The autoencoder is trained to reconstruct lesion images,and its latent embeddings are used as features for classification.To enhance minority-class representation,the Synthetic Minority Oversampling Technique(SMOTE)is applied directly to the latent vectors before classifier training.The encoder and classifier are first trained independently and later fine-tuned end-to-end.On the HAM10000 dataset,CANNSkin achieves an accuracy of 93.01%,a macro-F1 of 88.54%,and an ROC–AUC of 98.44%,demonstrating strong robustness across ten test subsets.Evaluation on the more complex ISIC 2019 dataset further confirms the model’s effectiveness,where CANNSkin achieves 94.27%accuracy,93.95%precision,94.09%recall,and 99.02%F1-score,supported by high reconstruction fidelity(PSNR 35.03 dB,SSIM 0.86).These results demonstrate the effectiveness of our proposed latent-space balancing and fine-tuned representation learning as a new benchmark method for robust and accurate skin cancer classification across heterogeneous datasets.
基金supported by the School of Digital Science,Universiti Brunei Darussalam,Brunei.
文摘Artificial Intelligence(AI)in healthcare enables predicting diabetes using data-driven methods instead of the traditional ways of screening the disease,which include hemoglobin A1c(HbA1c),oral glucose tolerance test(OGTT),and fasting plasma glucose(FPG)screening techniques,which are invasive and limited in scale.Machine learning(ML)and deep neural network(DNN)models that use large datasets to learn the complex,nonlinear feature interactions,but the conventional ML algorithms are data sensitive and often show unstable predictive accuracy.Conversely,DNN models are more robust,though the ability to reach a high accuracy rate consistently on heterogeneous datasets is still an open challenge.For predicting diabetes,this work proposed a hybrid DNN approach by integrating a bidirectional long short-term memory(BiLSTM)network with a bidirectional gated recurrent unit(BiGRU).A robust DL model,developed by combining various datasets with weighted coefficients,dense operations in the connection of deep layers,and the output aggregation using batch normalization and dropout functions to avoid overfitting.The goal of this hybrid model is better generalization and consistency among various datasets,which facilitates the effective management and early intervention.The proposed DNN model exhibits an excellent predictive performance as compared to the state-of-the-art and baseline ML and DNN models for diabetes prediction tasks.The robust performance indicates the possible usefulness of DL-based models in the development of disease prediction in healthcare and other areas that demand high-quality analytics.
基金supported by the National Natural Science Foundation of China (12473105 and 12473106)the central government guides local funds for science and technology development (YDZJSX2024D049)the Graduate Student Practice and Innovation Program of Shanxi Province (2024SJ313)
文摘As large-scale astronomical surveys,such as the Sloan Digital Sky Survey(SDSS)and the Large Sky Area Multi-Object Fiber Spectroscopic Telescope(LAMOST),generate increasingly complex datasets,clustering algorithms have become vital for identifying patterns and classifying celestial objects.This paper systematically investigates the application of five main categories of clustering techniques-partition-based,density-based,model-based,hierarchical,and“others”-across a range of astronomical research over the past decade.This review focuses on the six key application areas of stellar classification,galaxy structure analysis,detection of galactic and interstellar features,highenergy astrophysics,exoplanet studies,and anomaly detection.This paper provides an in-depth analysis of the performance and results of each method,considering their respective suitabilities for different data types.Additionally,it presents clustering algorithm selection strategies based on the characteristics of the spectroscopic data being analyzed.We highlight challenges such as handling large datasets,the need for more efficient computational tools,and the lack of labeled data.We also underscore the potential of unsupervised and semi-supervised clustering approaches to overcome these challenges,offering insight into their practical applications,performance,and results in astronomical research.
基金supported by grants from the Agriculture Seed Improvement Projects of Shandong Province(Nos.2022LZGCQY010,2021ZLGX03,and 2021TSGC 1240)the China Agriculture Research System Project(No.CARS-49)。
文摘The Chamidae are widely distributed in the tropical to temperate seas,with>70 known species.Currently,their classification relies mainly on traditional morphological methods and identification primarily on small fragment genes,such as COI.The intrafamily phylogenetic relationships are ambiguous,lacking support from reliable molecular data.In this study,the mitochondrial genomes of eight species of Chamidae were sequenced for the first time and then annotated.Their structures and compositional characteristics were analyzed.The mitochondrial gene order in this family differed significantly.Concurrently,the evolutionary position and phylogenetic relationship among Chamidae species were explored,and the Veneroida phylogenetic tree was recreated.Maximum Likelihood and Bayesian Inference analyses supported the monophyly of Chamidae.Additionally,the divergence time within Chamidae was estimated based on mitochondrial DNA sequences,with the most ancient divergence estimated to occur during the early Cretaceous period,128 MYA.This phylogeny is significant for understanding the diversity and taxonomic status of Chamidae.
基金funded by the National Natural Science Foundation of China(grant no.31670207).
文摘The genus Clematis(Ranunculaceae)comprises over 300 species with remarkable morphological and ecological diversity worldwide.Despite its horticultural,medicinal,and ecological importance,a well-resolved phylogeny and coherent infrageneric classification are still lacking.Here,we reconstruct a robust phylogeny for Clematis using a phylogenomic approach and revise its infrageneric taxonomy.We incorporated 198 samples representing 151 species,two subspecies,and 12 varieties,covering all subgenera and most sections worldwide,obtained from both fresh and herbarium material.Nuclear single nucleotide polymorphisms(SNPs)and complete plastid genomes were assembled for phylogenetic analyses.We also prepared a nuclear ribosomal ITS(nrITS)dataset comprising 171 species,two subspecies,and 12 varieties(217 samples)to include as many species as possible for phylogenetic inference.Phylogenies based on plastid genomes and nrITS exhibited limited resolution and modest support,highlighting challenges in resolving certain relationships.Nuclear SNP analyses yielded a robust phylogenetic tree with 22 well-supported clades corresponding to 22 sections,with most previously recognized subgenera and sections not recovered as monophyletic.Ancestral state reconstruction of 12 key morphological characters revealed multiple independent origins of character states.This study presents the first comprehensive sectional classification for Clematis based on robust phylogenomic evidence,redefines morphological characteristics for each section,and resolves long-standing taxonomic ambiguities.Our results establish a framework for future studies on the evolution,ecology,and horticultural potential of this globally significant genus.
文摘Classifying job offers into occupational categories is a fundamental task in human resource information systems,as it improves and streamlines indexing,search,and matching between openings and job seekers.Comprehensive occupational databases such as O∗NET or ESCO provide detailed taxonomies of interrelated positions that can be leveraged to align the textual content of postings with occupational categories,thereby facilitating standardization,cross-system interoperability,and access to metadata for each occupation(e.g.,tasks,knowledge,skills,and abilities).In this work,we explore the effectiveness of fine-tuning existing language models(LMs)to classify job offers with occupational descriptors from O∗NET.This enables a more precise assessment of candidate suitability by identifying the specific knowledge and skills required for each position,and helps automate recruitment processes by mitigating human bias and subjectivity in candidate selection.We evaluate three representative BERT-like models:BERT,RoBERTa,and DeBERTa.BERT serves as the baseline encoder-only architecture;RoBERTa incorporates advances in pretraining objectives and data scale;and DeBERTa introduces architectural improvements through disentangled attention mechanisms.The best performance was achieved with the DeBERTa model,although the other models also produced strong results,and no statistically significant differences were observed acrossmodels.We also find that these models typically reach optimal performance after only a few training epochs,and that training with smaller,balanced datasets is effective.Consequently,comparable results can be obtained with models that require fewer computational resources and less training time,facilitating deployment and practical use.
基金supported by the National Natural Science Foundation of China under Grant No.62162009the Key Technologies R&D Program of He’nan Province under Grant No.242102211065+2 种基金the Postgraduate Education Reform and Quality Improvement Project of Henan Province under Grant Nos.YJS2025GZZ36,YJS2024AL112,and YJS2024JD38the Innovation Scientists and Technicians Troop Construction Projects of Henan Province under Grant No.CXTD2017099the Scientific Research Innovation Team of Xuchang University under Grant No.2022CXTD003.
文摘With the increasing complexity of malware attack techniques,traditional detection methods face significant challenges,such as privacy preservation,data heterogeneity,and lacking category information.To address these issues,we propose Federated Dynamic Prototype Learning(FedDPL)for malware classification by integrating Federated Learning with a specifically designed K-means.Under the Federated Learning framework,model training occurs locally without data sharing,effectively protecting user data privacy and preventing the leakage of sensitive information.Furthermore,to tackle the challenges of data heterogeneity and the lack of category information,FedDPL introduces a dynamic prototype learning mechanism,which adaptively adjusts the clustering prototypes in terms of position and number.Thus,the dependency on predefined category numbers in typical K-means and its variants can be significantly reduced,resulting in improved clustering performance.Theoretically,it provides a more accurate detection of malicious behavior.Experimental results confirm that FedDPL excels in handling malware classification tasks,demonstrating superior accuracy,robustness,and privacy protection.
基金supported by the Innovative Human Resource Development for Local Intel-lectualization program through the Institute of Information&Communications Technology Planning&Evaluation(IITP)grant funded by the Korea government(MSIT)(No.IITP-2026-2020-0-01741)the research fund of Hanyang University(HY-2025-1110).
文摘Arrhythmias are a frequently occurring phenomenon in clinical practice,but how to accurately dis-tinguish subtle rhythm abnormalities remains an ongoing difficulty faced by the entire research community when conducting ECG-based studies.From a review of existing studies,two main factors appear to contribute to this problem:the uneven distribution of arrhythmia classes and the limited expressiveness of features learned by current models.To overcome these limitations,this study proposes a dual-path multimodal framework,termed DM-EHC(Dual-Path Multimodal ECG Heartbeat Classifier),for ECG-based heartbeat classification.The proposed framework links 1D ECG temporal features with 2D time–frequency features.By setting up the dual paths described above,the model can process more dimensions of feature information.The MIT-BIH arrhythmia database was selected as the baseline dataset for the experiments.Experimental results show that the proposed method outperforms single modalities and performs better for certain specific types of arrhythmias.The model achieved mean precision,recall,and F1 score of 95.14%,92.26%,and 93.65%,respectively.These results indicate that the framework is robust and has potential value in automated arrhythmia classification.
文摘The rapid digitalization of the energy sector has led to the deployment of large-scale smart metering systems that generate high-frequency time series data,creating new opportunities and challenges for energy anomaly detection.Accurate identification of anomalous patterns in building energy consumption is essential for optimizing operations,improving energy efficiency,and supporting grid reliability.This study investigates advanced feature engineering and machine learning modeling techniques for large-scale time series anomaly detection in building energy systems.Expanding upon previous benchmark frameworks,we introduce additional features such as oil price indices and solar cycle indicators,including sunset and sunrise times,to enhance the contextual understanding of consumption patterns.Our comparative modeling approach encompasses an extensive suite of algorithms,including KNeighborsUnif,KNeighborsDist,LightGBMXT,LightGBM,RandomForestMSE,CatBoost,ExtraTreesMSE,NeuralNetFastAI,XGBoost,NeuralNetTorch,and LightGBMLarge.Data preprocessing includes rigorous handling of missing values and normalization,while feature engineering focuses on temporal,environmental,and value-change attributes.The models are evaluated on a comprehensive dataset of smart meter readings,with performance assessed using metrics such as the Area Under the Receiver Operating Characteristic Curve(AUC-ROC).The results demonstrate that the integration of diverse exogenous variables and a hybrid ensemble of traditional tree-based and neural network models can significantly improve anomaly detection performance.This work provides new insights into the design of robust,scalable,and generalizable frameworks for energy anomaly detection in complex,real-world settings.
基金funded by Princess Nourah bint Abdulrahman University Researchers Supporting Project number(PNURSP2026R234)Princess Nourah bint Abdulrahman University,Riyadh,Saudi Arabia.
文摘Spam emails remain one of the most persistent threats to digital communication,necessitating effective detection solutions that safeguard both individuals and organisations.We propose a spam email classification frame-work that uses Bidirectional Encoder Representations from Transformers(BERT)for contextual feature extraction and a multiple-window Convolutional Neural Network(CNN)for classification.To identify semantic nuances in email content,BERT embeddings are used,and CNN filters extract discriminative n-gram patterns at various levels of detail,enabling accurate spam identification.The proposed model outperformed Word2Vec-based baselines on a sample of 5728 labelled emails,achieving an accuracy of 98.69%,AUC of 0.9981,F1 Score of 0.9724,and MCC of 0.9639.With a medium kernel size of(6,9)and compact multi-window CNN architectures,it improves performance.Cross-validation illustrates stability and generalization across folds.By balancing high recall with minimal false positives,our method provides a reliable and scalable solution for current spam detection in advanced deep learning.By combining contextual embedding and a neural architecture,this study develops a security analysis method.
基金supported by the Deanship of Scientific Research,Vice Presidency for Graduate Studies and Scientific Research,King Faisal University,Saudi Arabia[Grant No.KFU253774].
文摘Malware poses a significant threat to the Internet of Things(IoT).It enables unauthorized access to devices in the IoT environment.The lack of unique architectural standards causes challenges in developing robust malware detection(MD)models.The existing models demand substantial computational resources.This study intends to build a lightweight MD model to detect anomalies in IoT networks.The authors develop a transformation technique,converting the malware binaries into images.MobileNet V2 is fine-tuned using improved grey wolf optimization(IGWO)to extract crucial features of malicious and benign samples.The ResNeXt model is combined with the Linformer’s attention mechanism to identify Malware features.A fully connected layer is integrated with gradientweighted class activation mapping(Grad-CAM)in order to facilitate an interpretable classification model.The proposed model is evaluated using the IoT malware and the IoT-23 datasets.The model performs well on the two datasets with an accuracy of 98.94%,precision of 98.46%,recall of 98.11%,and F1-score of 98.28%on the IoT malware dataset,and an accuracy of 98.23%,precision of 96.80%,recall of 96.64%,and F1-score of 96.71%on the IoT-23 dataset,respectively.The findings indicate that the model has a high standard of classification.The lightweight architecture enables efficient deployment with an inference time of 1.42 s.Inference time has no direct impact on accuracy,precision,recall,or F1-score.However,the inference speed would warrant timely detection in latency-sensitive IoT applications.By achieving a remarkable result,the proposed study offers a comprehensive solution:a scalable,interpretable,and computationally efficient MD model for the evolving IoT landscape.
基金supported by Princess Nourah bint Abdulrahman University Researchers Supporting Project number(PNURSP2026R765),Princess Nourah bint Abdulrahman University,Riyadh,Saudi Arabia.
文摘Human Activity Recognition(HAR)is a novel area for computer vision.It has a great impact on healthcare,smart environments,and surveillance while is able to automatically detect human behavior.It plays a vital role in many applications,such as smart home,healthcare,human computer interaction,sports analysis,and especially,intelligent surveillance.In this paper,we propose a robust and efficient HAR system by leveraging deep learning paradigms,including pre-trained models,CNN architectures,and their average-weighted fusion.However,due to the diversity of human actions and various environmental influences,as well as a lack of data and resources,achieving high recognition accuracy remain elusive.In this work,a weighted average ensemble technique is employed to fuse three deep learning models:EfficientNet,ResNet50,and a custom CNN.The results of this study indicate that using a weighted average ensemble strategy for developing more effective HAR models may be a promising idea for detection and classification of human activities.Experiments by using the benchmark dataset proved that the proposed weighted ensemble approach outperformed existing approaches in terms of accuracy and other key performance measures.The combined average-weighted ensemble of pre-trained and CNN models obtained an accuracy of 98%,compared to 97%,96%,and 95%for the customized CNN,EfficientNet,and ResNet50 models,respectively.
基金supported by the Doctoral Foundation of Zunyi Normal University of China (BS [2022] 07)Fundação de Apoio à Pesquisa do Estado da Paraíba (FAPESQ)+5 种基金Conselho Nacional de Desenvolvimento Cientíıfico e Tecnológico (CNPq)-[150891/2023-7] for the financial supportsupported by the Q-CAYLE project, funded by the European Union-Next Generation UE/MICIU/Plan de Recuperacion, Transformacion y Resiliencia/Junta de Castilla y Leon (PRTRC17.11)by project PID2023-148409NB-I00, funded by MICIU/AEI/10.13039/501100011033Financial support of the Department of Education of the Junta de Castilla y LeonFEDER FundsExcellence project Fo S UHK 2203/2025-2026 for the financial support
文摘In this study,we explore a spherically symmetric charged black hole(BH)with a negative cosmological constant under the influence of a Kalb-Ramond field background.We compute the photon sphere and shadow radii,validating our findings using observational data from the Event Horizon Telescope,with a particular emphasis on the shadow images of Sagittarius A^(*).Furthermore,we investigate the greybody factors,emission rate,and partial absorption cross section.It is shown that the Lorentz-violating parameter l has an important effect on the absorption cross section.Our analysis also includes an examination of the topological charge,temperature-dependent topology,and generalized free energy.In particular,we regard the AdS charged BH with an antisymmetric tensor background as a topological defect in the thermodynamic space,then the system has the same topological classification to the charged Reissner-Nordström-AdS BH.
基金supported by the National Natural Science Foundation of China(NSFC)under Grant number:82171965.
文摘Designing appropriate loss functions is critical to the success of supervised learning models.However,most conventional losses are fixed and manually designed,making them suboptimal for diverse and dynamic learning scenarios.In this work,we propose an Adaptive Meta-Loss Network(Adaptive-MLN)that learns to generate taskagnostic loss functions tailored to evolving classification problems.Unlike traditional methods that rely on static objectives,Adaptive-MLN treats the loss function itself as a trainable component,parameterized by a shallow neural network.To enable flexible,gradient-free optimization,we introduce a hybrid evolutionary approach that combines GeneticAlgorithms(GA)for global exploration and Evolution Strategies(ES)for local refinement.This co-evolutionary process dynamically adjusts the loss landscape,improvingmodel generalization without relying on analytic gradients or handcrafted heuristics.Experimental evaluations on synthetic tasks and the CIFAR-10 andMNIST datasets demonstrate that our approach consistently outperforms standard losses such as Cross-Entropy and Mean Squared Error in terms of accuracy,convergence,and adaptability.
基金supported by the National Natural Science Foundation of China(42201339)the“Science for a Better Development of Inner Mongolia”Program of the Bureau of Science and Technology of the Inner Mongolia Autonomous Region,China(2022EEDSKJXM003)。
文摘The fine-scale characterization of vegetation surface information serves as a fundamental basis for studying the spatial distribution of resources and the dynamic patterns of environmental responses.Accurately extracting the distributions of different crop species is of critical importance for improving agricultural production efficiency and ensuring food security.Traditional fine-scale vegetation extraction methods often face significant challenges due to the presence of spectrally similar features and the substantial influence of background interference,which limit their applicability across large areas.As a key phenological stage of angiosperms,flowering is characterized by distinctive flowering times,floral morphology,and canopy spectral signatures,so it is an effective pathway for fine-scale vegetation extraction using remote sensing.Using rapeseed as an example,this study developed a spectral index model for precise flowering vegetation extraction(FI-R)based on Landsat OLI imagery.The model integrates a yellowness index(Blue,Green)and a peak index(Red,Nir and SWIR1)while leveraging the NDVI to mitigate background interference from spectrally similar objects.This approach successfully enables the rapid and accurate large-scale mapping of flowering vegetation under complex background conditions.The proposed method was tested in five rapeseed cultivation regions worldwide with diverse backgrounds.Validation datasets were generated using GF imagery and the U.S.CDL dataset.The FI-R model demonstrated superior capability in distinguishing flowering rapeseed from other vegetation,and achieved overall accuracies exceeding 94%in all study areas.Furthermore,FI-R is compatible with other multispectral sensors that have similar band configurations,so it is applicable to rapeseed extraction in broader contexts.The method also shows strong potential for the fine-scale extraction of other types of flowering angiosperm vegetation.
基金supported by the SungKyunKwan University and the BK21 FOUR(Graduate School Innovation)funded by the Ministry of Education(MOE,Korea)and National Research Foundation of Korea(NRF).
文摘With the recent increase in data volume and diversity,traditional text representation techniques are struggling to capture context,particularly in environments with sparse data.To address these challenges,this study proposes a new model,the Masked Joint Representation Model(MJRM).MJRM approximates the original hypothesis by leveraging multiple elements in a limited context.It dynamically adapts to changes in characteristics based on data distribution through three main components.First,masking-based representation learning,termed selective dynamic masking,integrates topic modeling and sentiment clustering to generate and train multiple instances across different data subsets,whose predictions are then aggregated with optimized weights.This design alleviates sparsity,suppresses noise,and preserves contextual structures.Second,regularization-based improvements are applied.Third,techniques for addressing sparse data are used to perform final inference.As a result,MJRM improves performance by up to 4%compared to existing AI techniques.In our experiments,we analyzed the contribution of each factor,demonstrating that masking,dynamic learning,and aggregating multiple instances complement each other to improve performance.This demonstrates that a masking-based multi-learning strategy is effective for context-aware sparse text classification,and can be useful even in challenging situations such as data shortage or data distribution variations.We expect that the approach can be extended to diverse fields such as sentiment analysis,spam filtering,and domain-specific document classification.
基金supported by the National Natural Science Foundation of China(U2344201 and 42101316)the Natural Science Foundation of Hunan Province,China(2022JJ40866)the Outstanding Youth Project of Education Bureau of Hunan Province,China(20B613)。
文摘Understanding the spatial distributions and corresponding variation mechanisms of key soil nutrients in fragile karst ecosystems can assist in promoting sustainable development.However,due to the implementation of ecological restoration initiatives such as land-use conversions,novel changes in the spatial characteristics of soil nutrients remain unknown.To address this gap,we explored nutrient variations and the drivers of the variation in the 0–15 cm topsoil layer using a regional-scale sampling method in a typical karst area in northwest Guangxi Zhuang Autonomous Region,Southwest China.Descriptive statistics,geostatistics,and spatial analysis were used to assess the soil nutrient variability.The results indicated that soil organic carbon(SOC),total nitrogen(TN),total phosphorus(TP),and total potassium(TK)concentrations showed moderate variations,with coefficients of variance being 0.60,0.60,0.71,and 0.72,respectively.Moreover,they demonstrated positive spatial autocorrelations,with global Moran's indices being 0.68,0.77,0.64,and 0.68,respectively.However,local Moran's index values were low,indicating large spatial variations in soil nutrients.The best-fitting semi-variogram models for SOC,TN,TP,and TK concentrations were spherical,Gaussian,exponential,and exponential,respectively.According to the classification criteria of the Second National Soil Census in China,SOC and TN concentrations were relatively sufficient,with the proportions of rich and very rich levels being up to 90.9 and 96.0%,respectively.TP concentration was in the mediumdeficient level,with the areas of medium and deficient levels accounting for 33.7 and 30.1%of the total,respectively.TK concentration was deficient,with the cumulative area of extremely deficient,very deficient,and deficient levels accounting for 87.6%of the total area.Consequently,the terrestrial ecosystems in the study area were more vulnerable to soil P and K than soil N deficiencies.Furthermore,variance partitioning analysis of the influencing factors showed that,except for the interactions,the single effect of other soil properties accounted more for soil nutrient variations than spatial and environmental variables.These results will aid in the future management of terrestrial ecosystems.
