In response to the scarcity of infrared aircraft samples and the tendency of traditional deep learning to overfit,a few-shot infrared aircraft classification method based on cross-correlation networks is proposed.This...In response to the scarcity of infrared aircraft samples and the tendency of traditional deep learning to overfit,a few-shot infrared aircraft classification method based on cross-correlation networks is proposed.This method combines two core modules:a simple parameter-free self-attention and cross-attention.By analyzing the self-correlation and cross-correlation between support images and query images,it achieves effective classification of infrared aircraft under few-shot conditions.The proposed cross-correlation network integrates these two modules and is trained in an end-to-end manner.The simple parameter-free self-attention is responsible for extracting the internal structure of the image while the cross-attention can calculate the cross-correlation between images further extracting and fusing the features between images.Compared with existing few-shot infrared target classification models,this model focuses on the geometric structure and thermal texture information of infrared images by modeling the semantic relevance between the features of the support set and query set,thus better attending to the target objects.Experimental results show that this method outperforms existing infrared aircraft classification methods in various classification tasks,with the highest classification accuracy improvement exceeding 3%.In addition,ablation experiments and comparative experiments also prove the effectiveness of the method.展开更多
Image classification is crucial for various applications,including digital construction,smart manu-facturing,and medical imaging.Focusing on the inadequate model generalization and data privacy concerns in few-shot im...Image classification is crucial for various applications,including digital construction,smart manu-facturing,and medical imaging.Focusing on the inadequate model generalization and data privacy concerns in few-shot image classification,in this paper,we propose a federated learning approach that incorporates privacy-preserving techniques.First,we utilize contrastive learning to train on local few-shot image data and apply various data augmentation methods to expand the sample size,thereby enhancing the model’s generalization capabilities in few-shot contexts.Second,we introduce local differential privacy techniques and weight pruning methods to safeguard model parameters,perturbing the transmitted parameters to ensure user data privacy.Finally,numerical simulations are conducted to demonstrate the effectiveness of our proposed method.The results indicate that our approach significantly enhances model generalization and test accuracy compared to several popular federated learning algorithms while maintaining data privacy,highlighting its effectiveness and practicality in addressing the challenges of model generalization and data privacy in few-shot image scenarios.展开更多
Existing weakly supervised semantic segmentation(WSSS)methods based on image-level labels always rely on class activation maps(CAMs),which measure the relationships between features and classifiers.However,CAMs only f...Existing weakly supervised semantic segmentation(WSSS)methods based on image-level labels always rely on class activation maps(CAMs),which measure the relationships between features and classifiers.However,CAMs only focus on the most discriminative regions of images,resulting in their poor coverage performance.We attribute this to the deficiency in the recognition ability of a single classifier and the negative impacts caused by magnitudes during the CAMs normalisation process.To address the aforementioned issues,we propose to construct selective multiple classifiers(SMC).During the training process,we extract multiple prototypes for each class and store them in the corresponding memory bank.These prototypes are divided into foreground and background prototypes,with the former used to identify foreground objects and the latter aimed at preventing the false activation of background pixels.As for the inference stage,multiple prototypes are adaptively selected from the memory bank for each image as SMC.Subsequently,CAMs are generated by measuring the angle between SMC and features.We enhance the recognition ability of classifiers by adaptively constructing multiple classifiers for each image,while only relying on angle measurement to generate CAMs can alleviate the suppression phenomenon caused by magnitudes.Furthermore,SMC can be integrated into other WSSS approaches to help generate better CAMs.Extensive experiments conducted on standard WSSS benchmarks such as PASCAL VOC 2012 and MS COCO 2014 demonstrate the superiority of our proposed method.展开更多
Human Activity Recognition(HAR)in drone-captured videos has become popular because of the interest in various fields such as video surveillance,sports analysis,and human-robot interaction.However,recognizing actions f...Human Activity Recognition(HAR)in drone-captured videos has become popular because of the interest in various fields such as video surveillance,sports analysis,and human-robot interaction.However,recognizing actions from such videos poses the following challenges:variations of human motion,the complexity of backdrops,motion blurs,occlusions,and restricted camera angles.This research presents a human activity recognition system to address these challenges by working with drones’red-green-blue(RGB)videos.The first step in the proposed system involves partitioning videos into frames and then using bilateral filtering to improve the quality of object foregrounds while reducing background interference before converting from RGB to grayscale images.The YOLO(You Only Look Once)algorithm detects and extracts humans from each frame,obtaining their skeletons for further processing.The joint angles,displacement and velocity,histogram of oriented gradients(HOG),3D points,and geodesic Distance are included.These features are optimized using Quadratic Discriminant Analysis(QDA)and utilized in a Neuro-Fuzzy Classifier(NFC)for activity classification.Real-world evaluations on the Drone-Action,Unmanned Aerial Vehicle(UAV)-Gesture,and Okutama-Action datasets substantiate the proposed system’s superiority in accuracy rates over existing methods.In particular,the system obtains recognition rates of 93%for drone action,97%for UAV gestures,and 81%for Okutama-action,demonstrating the system’s reliability and ability to learn human activity from drone videos.展开更多
Background:In the field of genetic diagnostics,DNA sequencing is an important tool because the depth and complexity of this field have major implications in light of the genetic architectures of diseases and the ident...Background:In the field of genetic diagnostics,DNA sequencing is an important tool because the depth and complexity of this field have major implications in light of the genetic architectures of diseases and the identification of risk factors associated with genetic disorders.Methods:Our study introduces a novel two-tiered analytical framework to raise the precision and reliability of genetic data interpretation.It is initiated by extracting and analyzing salient features from DNA sequences through a CNN-based feature analysis,taking advantage of the power inherent in Convolutional neural networks(CNNs)to attain complex patterns and minute mutations in genetic data.This study embraces an elite collection of machine learning classifiers interweaved through a stern voting mechanism,which synergistically joins the predictions made from multiple classifiers to generate comprehensive and well-balanced interpretations of the genetic data.Results:This state-of-the-art method was further tested by carrying out an empirical analysis on a variants'dataset of DNA sequences taken from patients affected by breast cancer,juxtaposed with a control group composed of healthy people.Thus,the integration of CNNs with a voting-based ensemble of classifiers returned outstanding outcomes,with performance metrics accuracy,precision,recall,and F1-scorereaching the outstanding rate of 0.88,outperforming previous models.Conclusions:This dual accomplishment underlines the transformative potential that integrating deep learning techniques with ensemble machine learning might provide in real added value for further genetic diagnostics and prognostics.These results from this study set a new benchmark in the accuracy of disease diagnosis through DNA sequencing and promise future studies on improved personalized medicine and healthcare approaches with precise genetic information.展开更多
Background:Stomach cancer(SC)is one of the most lethal malignancies worldwide due to late-stage diagnosis and limited treatment.The transcriptomic,epigenomic,and proteomic,etc.,omics datasets generated by high-through...Background:Stomach cancer(SC)is one of the most lethal malignancies worldwide due to late-stage diagnosis and limited treatment.The transcriptomic,epigenomic,and proteomic,etc.,omics datasets generated by high-throughput sequencing technology have become prominent in biomedical research,and they reveal molecular aspects of cancer diagnosis and therapy.Despite the development of advanced sequencing technology,the presence of high-dimensionality in multi-omics data makes it challenging to interpret the data.Methods:In this study,we introduce RankXLAN,an explainable ensemble-based multi-omics framework that integrates feature selection(FS),ensemble learning,bioinformatics,and in-silico validation for robust biomarker detection,potential therapeutic drug-repurposing candidates’identification,and classification of SC.To enhance the interpretability of the model,we incorporated explainable artificial intelligence(SHapley Additive exPlanations analysis),as well as accuracy,precision,F1-score,recall,cross-validation,specificity,likelihood ratio(LR)+,LR−,and Youden index results.Results:The experimental results showed that the top four FS algorithms achieved improved results when applied to the ensemble learning classification model.The proposed ensemble model produced an area under the curve(AUC)score of 0.994 for gene expression,0.97 for methylation,and 0.96 for miRNA expression data.Through the integration of bioinformatics and ML approach of the transcriptomic and epigenomic multi-omics dataset,we identified potential marker genes,namely,UBE2D2,HPCAL4,IGHA1,DPT,and FN3K.In-silico molecular docking revealed a strong binding affinity between ANKRD13C and the FDA-approved drug Everolimus(binding affinity−10.1 kcal/mol),identifying ANKRD13C as a potential therapeutic drug-repurposing target for SC.Conclusion:The proposed framework RankXLAN outperforms other existing frameworks for serum biomarker identification,therapeutic target identification,and SC classification with multi-omics datasets.展开更多
To improve the performance of the multiple classifier system, a new method of feature-decision level fusion is proposed based on knowledge discovery. In the new method, the base classifiers operate on different featur...To improve the performance of the multiple classifier system, a new method of feature-decision level fusion is proposed based on knowledge discovery. In the new method, the base classifiers operate on different feature spaces and their types depend on different measures of between-class separability. The uncertainty measures corresponding to each output of each base classifier are induced from the established decision tables （DTs） in the form of mass function in the Dempster-Shafer theory （DST）. Furthermore, an effective fusion framework is built at the feature-decision level on the basis of a generalized rough set model and the DST. The experiment for the classification of hyperspectral remote sensing images shows that the performance of the classification can be improved by the proposed method compared with that of plurality voting （PV）.展开更多
To improve the performance of multiple classifier system, a knowledge discovery based dynamic weighted voting （KD-DWV） is proposed based on knowledge discovery. In the method, all base classifiers may be allowed to ...To improve the performance of multiple classifier system, a knowledge discovery based dynamic weighted voting （KD-DWV） is proposed based on knowledge discovery. In the method, all base classifiers may be allowed to operate in different measurement/feature spaces to make the most of diverse classification information. The weights assigned to each output of a base classifier are estimated by the separability of training sample sets in relevant feature space. For this purpose, some decision tables （DTs） are established in terms of the diverse feature sets. And then the uncertainty measures of the separability are induced, in the form of mass functions in Dempster-Shafer theory （DST）, from each DTs based on generalized rough set model. From the mass functions, all the weights are calculated by a modified heuristic fusion function and assigned dynamically to each classifier varying with its output. The comparison experiment is performed on the hyperspectral remote sensing images. And the experimental results show that the performance of the classification can be improved by using the proposed method compared with the plurality voting （PV）.展开更多
Support vector machine (SVM), as a novel approach in pattern recognition, has demonstrated a success in face detection and face recognition. In this paper, a face recognition approach based on the SVM classifier with ...Support vector machine (SVM), as a novel approach in pattern recognition, has demonstrated a success in face detection and face recognition. In this paper, a face recognition approach based on the SVM classifier with the nearest neighbor classifier (NNC) is proposed. The principal component analysis (PCA) is used to reduce the dimension and extract features. Then one-against-all stratedy is used to train the SVM classifiers. At the testing stage, we propose an al-展开更多
Numerous models have been proposed to reduce the classification error of Naive Bayes by weakening its attribute independence assumption and some have demonstrated remarkable error performance. Considering that ensembl...Numerous models have been proposed to reduce the classification error of Naive Bayes by weakening its attribute independence assumption and some have demonstrated remarkable error performance. Considering that ensemble learning is an effective method of reducing the classifmation error of the classifier, this paper proposes a double-layer Bayesian classifier ensembles （DLBCE） algorithm based on frequent itemsets. DLBCE constructs a double-layer Bayesian classifier （DLBC） for each frequent itemset the new instance contained and finally ensembles all the classifiers by assigning different weight to different classifier according to the conditional mutual information. The experimental results show that the proposed algorithm outperforms other outstanding algorithms.展开更多
This paper proposed an algorithm in which the maximum probability and the weighted average strategy were used for the combination of member classifiers. Using parallel computing, we test the algorithm on a China-Brazi...This paper proposed an algorithm in which the maximum probability and the weighted average strategy were used for the combination of member classifiers. Using parallel computing, we test the algorithm on a China-Brazil Earth Resources Satellite （CBERS） image for land cover classification. The results show that using three computers in parallel can reduce the classification time by 30%, as compared with using only one computer with a dual core processor. The accuracy of the final image is 93.34%, and Kappa is 0.92. Multiple classifier combination can enhance the precision of the image classification, and parallel computing can increase the speed of calculation so that it becomes possible to process remote sensing images with high efficiency and accuracy.展开更多
The participation of ordinary devices in networking has created a world of connected devices rapidly.The Internet of Things(IoT)includes heterogeneous devices from every field.There are no definite protocols or standa...The participation of ordinary devices in networking has created a world of connected devices rapidly.The Internet of Things(IoT)includes heterogeneous devices from every field.There are no definite protocols or standards for IoT communication,and most of the IoT devices have limited resources.Enabling a complete security measure for such devices is a challenging task,yet necessary.Many lightweight security solutions have surfaced lately for IoT.The lightweight security protocols are unable to provide an optimum protection against prevailing powerful threats in cyber world.It is also hard to deploy any traditional security protocol on resource-constrained IoT devices.Software-defined networking introduces a centralized control in computer networks.SDN has a programmable approach towards networking that decouples control and data planes.An SDN-based intrusion detection system is proposed which uses deep learning classifier for detection of anomalies in IoT.The proposed intrusion detection system does not burden the IoT devices with security profiles.The proposed work is executed on the simulated environment.The results of the simulation test are evaluated using various matrices and compared with other relevant methods.展开更多
Mapping and monitoring the distribution of croplands and crop types support policymakers and international organizations by reducing the risks to food security,notably from climate change and,for that purpose,remote s...Mapping and monitoring the distribution of croplands and crop types support policymakers and international organizations by reducing the risks to food security,notably from climate change and,for that purpose,remote sensing is routinely used.However,identifying specific crop types,cropland,and cropping patterns using space-based observations is challenging because different crop types and cropping patterns have similarity spectral signatures.This study applied a methodology to identify cropland and specific crop types,including tobacco,wheat,barley,and gram,as well as the following cropping patterns:wheat-tobacco,wheat-gram,wheat-barley,and wheat-maize,which are common in Gujranwala District,Pakistan,the study region.The methodology consists of combining optical remote sensing images from Sentinel-2 and Landsat-8 with Machine Learning(ML)methods,namely a Decision Tree Classifier(DTC)and a Random Forest(RF)algorithm.The best time-periods for differentiating cropland from other land cover types were identified,and then Sentinel-2 and Landsat 8 NDVI-based time-series were linked to phenological parameters to determine the different crop types and cropping patterns over the study region using their temporal indices and ML algorithms.The methodology was subsequently evaluated using Landsat images,crop statistical data for 2020 and 2021,and field data on cropping patterns.The results highlight the high level of accuracy of the methodological approach presented using Sentinel-2 and Landsat-8 images,together with ML techniques,for mapping not only the distribution of cropland,but also crop types and cropping patterns when validated at the county level.These results reveal that this methodology has benefits for monitoring and evaluating food security in Pakistan,adding to the evidence base of other studies on the use of remote sensing to identify crop types and cropping patterns in other countries.展开更多
Deep learning has been fully verified and accepted in the field of electromagnetic signal classification. However, in many specific scenarios, such as radio resource management for aircraft communications, labeled dat...Deep learning has been fully verified and accepted in the field of electromagnetic signal classification. However, in many specific scenarios, such as radio resource management for aircraft communications, labeled data are difficult to obtain, which makes the best deep learning methods at present seem almost powerless, because these methods need a large amount of labeled data for training. When the training dataset is small, it is highly possible to fall into overfitting, which causes performance degradation of the deep neural network. For few-shot electromagnetic signal classification, data augmentation is one of the most intuitive countermeasures. In this work, a generative adversarial network based on the data augmentation method is proposed to achieve better classification performance for electromagnetic signals. Based on the similarity principle, a screening mechanism is established to obtain high-quality generated signals. Then, a data union augmentation algorithm is designed by introducing spatiotemporally flipped shapes of the signal. To verify the effectiveness of the proposed data augmentation algorithm, experiments are conducted on the RADIOML 2016.04C dataset and real-world ACARS dataset. The experimental results show that the proposed method significantly improves the performance of few-shot electromagnetic signal classification.展开更多
The turbo air classifier is widely used powder classification equipment in a variety of fields. The flow field characteristics of the turbo air classifier are important basis for the improvement of the turbo air class...The turbo air classifier is widely used powder classification equipment in a variety of fields. The flow field characteristics of the turbo air classifier are important basis for the improvement of the turbo air classifier＇s structural design. The flow field characteristics of the rotor cage in turbo air classifiers were investigated trader different operating conditions by laser Doppler velocimeter（LDV）, and a measure diminishing the axial velocity is proposed. The investigation results show that the tangential velocity of the air flow inside the rotor cage is different from the rotary speed of the rotor cage on the same measurement point due to the influences of both the negative pressure at the exit and the rotation of the rotor cage. The tangential velocity of the air flow likewise decreases as the radius decreases in the case of the rotor cage＇s low rotary speed. In contrast, the tangential velocity of the air flow increases as the radius decreases in the case of the rotor cage＇s high rotary speed. Meanwhile, the vortex inside the rotor cage is found to occur near the pressure side of the blade when the rotor cage＇s rotary speed is less than the tangential velocity of air flow. On the contrary, the vortex is found to occur near the blade suction side once the rotor cage＇s rotary speed is higher than the tangential velocity of air flow. Inside the rotor cage, the axial velocity could not be disregarded and is largely determined by the distances between the measurement point and the exit.展开更多
In this work,the reflux classifier with closely spaced inclined channels is used as the pre-concentration facility to improve the separation efficiency before the shaking table separation.Three operating parameters of...In this work,the reflux classifier with closely spaced inclined channels is used as the pre-concentration facility to improve the separation efficiency before the shaking table separation.Three operating parameters of reflux classifier(RC)to pre-concentrate fine(0.023−0.15 mm)tailings of antimony oxide were optimized by response surface methodology(RSM)using a three-level Box-Behnken design(BBD).The parameters studied for the optimization were feeding speed,underflow,and ascending water speed.Second-order response functions were produced for the Sb grade and recovery rate of the concentrate.Taking advantage of the quadratic programming,when the factors of feeding,underflow and ascending water are respectively 225,30 and 133 cm^3/min,a better result can be achieved for the concentrate grade of 2.31% and recovery rate of 83.17%.At the same time,70.48% of the tailings with the grade of 0.20% were discarded out of the feeding.The results indicated that the reflux classifier has a good performance in dealing with fine tailings of antimony oxide.Moreover,second-order polynomial equations,ANOVA,and three-dimensional surface plots were developed to evaluate the effects of each parameter on Sb grade and recovery rate of the concentrate.展开更多
Predicting stock price movements is a challenging task for academicians and practitioners. In particular, forecasting price movements in emerging markets seems to be more elusive because they are usually more volatile...Predicting stock price movements is a challenging task for academicians and practitioners. In particular, forecasting price movements in emerging markets seems to be more elusive because they are usually more volatile often accompa-nied by thin trading-volumes and they are susceptible to more manipulation compared to mature markets. Technical analysis of stocks and commodities has become a science on its own;quantitative methods and techniques have been applied by many practitioners to forecast price movements. Lagging and sometimes leading technical indicators pro-vide rich quantitative tools for traders and investors in their attempt to gain advantage when making investment or trading decisions. Artificial Neural Networks (ANN) have been used widely in predicting stock prices because of their capability in capturing the non-linearity that often exists in price movements. Recently, Polynomial Classifiers (PC) have been applied to various recognition and classification application and showed favorable results in terms of recog-nition rates and computational complexity as compared to ANN. In this paper, we present two prediction models for predicting securities’ prices. The first model was developed using back propagation feed forward neural networks. The second model was developed using polynomial classifiers (PC), as a first time application for PC to be used in stock prices prediction. The inputs to both models were identical, and both models were trained and tested on the same data. The study was conducted on Dubai Financial Market as an emerging market and applied to two of the market’s leading stocks. In general, both models achieved very good results in terms of mean absolute error percentage. Both models show an average error around 1.5% predicting the next day price, an average error of 2.5% when predicting second day price, and an average error of 4% when predicted the third day price.展开更多
基金Supported by the National Pre-research Program during the 14th Five-Year Plan(514010405)。
文摘In response to the scarcity of infrared aircraft samples and the tendency of traditional deep learning to overfit,a few-shot infrared aircraft classification method based on cross-correlation networks is proposed.This method combines two core modules:a simple parameter-free self-attention and cross-attention.By analyzing the self-correlation and cross-correlation between support images and query images,it achieves effective classification of infrared aircraft under few-shot conditions.The proposed cross-correlation network integrates these two modules and is trained in an end-to-end manner.The simple parameter-free self-attention is responsible for extracting the internal structure of the image while the cross-attention can calculate the cross-correlation between images further extracting and fusing the features between images.Compared with existing few-shot infrared target classification models,this model focuses on the geometric structure and thermal texture information of infrared images by modeling the semantic relevance between the features of the support set and query set,thus better attending to the target objects.Experimental results show that this method outperforms existing infrared aircraft classification methods in various classification tasks,with the highest classification accuracy improvement exceeding 3%.In addition,ablation experiments and comparative experiments also prove the effectiveness of the method.
基金supported by Suzhou Science and Technology Plan(Basic Research)Project under Grant SJC2023002Postgraduate Research&Practice Innovation Program of Jiangsu Province under Grant KYCX23_3322.
文摘Image classification is crucial for various applications,including digital construction,smart manu-facturing,and medical imaging.Focusing on the inadequate model generalization and data privacy concerns in few-shot image classification,in this paper,we propose a federated learning approach that incorporates privacy-preserving techniques.First,we utilize contrastive learning to train on local few-shot image data and apply various data augmentation methods to expand the sample size,thereby enhancing the model’s generalization capabilities in few-shot contexts.Second,we introduce local differential privacy techniques and weight pruning methods to safeguard model parameters,perturbing the transmitted parameters to ensure user data privacy.Finally,numerical simulations are conducted to demonstrate the effectiveness of our proposed method.The results indicate that our approach significantly enhances model generalization and test accuracy compared to several popular federated learning algorithms while maintaining data privacy,highlighting its effectiveness and practicality in addressing the challenges of model generalization and data privacy in few-shot image scenarios.
基金supported by the National Natural Science Foundation of China(Grants 62176097,61433007)Fundamental Research Funds for the Central Universities(Grant 2019kfyXKJC024)the 111 Project on Computational Intelligence and Intelligent Control(Grant B18024).
文摘Existing weakly supervised semantic segmentation(WSSS)methods based on image-level labels always rely on class activation maps(CAMs),which measure the relationships between features and classifiers.However,CAMs only focus on the most discriminative regions of images,resulting in their poor coverage performance.We attribute this to the deficiency in the recognition ability of a single classifier and the negative impacts caused by magnitudes during the CAMs normalisation process.To address the aforementioned issues,we propose to construct selective multiple classifiers(SMC).During the training process,we extract multiple prototypes for each class and store them in the corresponding memory bank.These prototypes are divided into foreground and background prototypes,with the former used to identify foreground objects and the latter aimed at preventing the false activation of background pixels.As for the inference stage,multiple prototypes are adaptively selected from the memory bank for each image as SMC.Subsequently,CAMs are generated by measuring the angle between SMC and features.We enhance the recognition ability of classifiers by adaptively constructing multiple classifiers for each image,while only relying on angle measurement to generate CAMs can alleviate the suppression phenomenon caused by magnitudes.Furthermore,SMC can be integrated into other WSSS approaches to help generate better CAMs.Extensive experiments conducted on standard WSSS benchmarks such as PASCAL VOC 2012 and MS COCO 2014 demonstrate the superiority of our proposed method.
基金funded by the Open Access Initiative of the University of Bremen and the DFG via SuUB Bremen.Princess Nourah bint Abdulrahman University Researchers Supporting Project number(PNURSP2024R348),Princess Nourah bint Abdulrahman University,Riyadh,Saudi Arabia.
文摘Human Activity Recognition(HAR)in drone-captured videos has become popular because of the interest in various fields such as video surveillance,sports analysis,and human-robot interaction.However,recognizing actions from such videos poses the following challenges:variations of human motion,the complexity of backdrops,motion blurs,occlusions,and restricted camera angles.This research presents a human activity recognition system to address these challenges by working with drones’red-green-blue(RGB)videos.The first step in the proposed system involves partitioning videos into frames and then using bilateral filtering to improve the quality of object foregrounds while reducing background interference before converting from RGB to grayscale images.The YOLO(You Only Look Once)algorithm detects and extracts humans from each frame,obtaining their skeletons for further processing.The joint angles,displacement and velocity,histogram of oriented gradients(HOG),3D points,and geodesic Distance are included.These features are optimized using Quadratic Discriminant Analysis(QDA)and utilized in a Neuro-Fuzzy Classifier(NFC)for activity classification.Real-world evaluations on the Drone-Action,Unmanned Aerial Vehicle(UAV)-Gesture,and Okutama-Action datasets substantiate the proposed system’s superiority in accuracy rates over existing methods.In particular,the system obtains recognition rates of 93%for drone action,97%for UAV gestures,and 81%for Okutama-action,demonstrating the system’s reliability and ability to learn human activity from drone videos.
文摘Background:In the field of genetic diagnostics,DNA sequencing is an important tool because the depth and complexity of this field have major implications in light of the genetic architectures of diseases and the identification of risk factors associated with genetic disorders.Methods:Our study introduces a novel two-tiered analytical framework to raise the precision and reliability of genetic data interpretation.It is initiated by extracting and analyzing salient features from DNA sequences through a CNN-based feature analysis,taking advantage of the power inherent in Convolutional neural networks(CNNs)to attain complex patterns and minute mutations in genetic data.This study embraces an elite collection of machine learning classifiers interweaved through a stern voting mechanism,which synergistically joins the predictions made from multiple classifiers to generate comprehensive and well-balanced interpretations of the genetic data.Results:This state-of-the-art method was further tested by carrying out an empirical analysis on a variants'dataset of DNA sequences taken from patients affected by breast cancer,juxtaposed with a control group composed of healthy people.Thus,the integration of CNNs with a voting-based ensemble of classifiers returned outstanding outcomes,with performance metrics accuracy,precision,recall,and F1-scorereaching the outstanding rate of 0.88,outperforming previous models.Conclusions:This dual accomplishment underlines the transformative potential that integrating deep learning techniques with ensemble machine learning might provide in real added value for further genetic diagnostics and prognostics.These results from this study set a new benchmark in the accuracy of disease diagnosis through DNA sequencing and promise future studies on improved personalized medicine and healthcare approaches with precise genetic information.
基金the Deanship of Research and Graduate Studies at King Khalid University,KSA,for funding this work through the Large Research Project under grant number RGP2/164/46.
文摘Background:Stomach cancer(SC)is one of the most lethal malignancies worldwide due to late-stage diagnosis and limited treatment.The transcriptomic,epigenomic,and proteomic,etc.,omics datasets generated by high-throughput sequencing technology have become prominent in biomedical research,and they reveal molecular aspects of cancer diagnosis and therapy.Despite the development of advanced sequencing technology,the presence of high-dimensionality in multi-omics data makes it challenging to interpret the data.Methods:In this study,we introduce RankXLAN,an explainable ensemble-based multi-omics framework that integrates feature selection(FS),ensemble learning,bioinformatics,and in-silico validation for robust biomarker detection,potential therapeutic drug-repurposing candidates’identification,and classification of SC.To enhance the interpretability of the model,we incorporated explainable artificial intelligence(SHapley Additive exPlanations analysis),as well as accuracy,precision,F1-score,recall,cross-validation,specificity,likelihood ratio(LR)+,LR−,and Youden index results.Results:The experimental results showed that the top four FS algorithms achieved improved results when applied to the ensemble learning classification model.The proposed ensemble model produced an area under the curve(AUC)score of 0.994 for gene expression,0.97 for methylation,and 0.96 for miRNA expression data.Through the integration of bioinformatics and ML approach of the transcriptomic and epigenomic multi-omics dataset,we identified potential marker genes,namely,UBE2D2,HPCAL4,IGHA1,DPT,and FN3K.In-silico molecular docking revealed a strong binding affinity between ANKRD13C and the FDA-approved drug Everolimus(binding affinity−10.1 kcal/mol),identifying ANKRD13C as a potential therapeutic drug-repurposing target for SC.Conclusion:The proposed framework RankXLAN outperforms other existing frameworks for serum biomarker identification,therapeutic target identification,and SC classification with multi-omics datasets.
文摘To improve the performance of the multiple classifier system, a new method of feature-decision level fusion is proposed based on knowledge discovery. In the new method, the base classifiers operate on different feature spaces and their types depend on different measures of between-class separability. The uncertainty measures corresponding to each output of each base classifier are induced from the established decision tables （DTs） in the form of mass function in the Dempster-Shafer theory （DST）. Furthermore, an effective fusion framework is built at the feature-decision level on the basis of a generalized rough set model and the DST. The experiment for the classification of hyperspectral remote sensing images shows that the performance of the classification can be improved by the proposed method compared with that of plurality voting （PV）.
基金This project was supported by the National Basic Research Programof China (2001CB309403)
文摘To improve the performance of multiple classifier system, a knowledge discovery based dynamic weighted voting （KD-DWV） is proposed based on knowledge discovery. In the method, all base classifiers may be allowed to operate in different measurement/feature spaces to make the most of diverse classification information. The weights assigned to each output of a base classifier are estimated by the separability of training sample sets in relevant feature space. For this purpose, some decision tables （DTs） are established in terms of the diverse feature sets. And then the uncertainty measures of the separability are induced, in the form of mass functions in Dempster-Shafer theory （DST）, from each DTs based on generalized rough set model. From the mass functions, all the weights are calculated by a modified heuristic fusion function and assigned dynamically to each classifier varying with its output. The comparison experiment is performed on the hyperspectral remote sensing images. And the experimental results show that the performance of the classification can be improved by using the proposed method compared with the plurality voting （PV）.
基金This project was supported by Shanghai Shu Guang Project.
文摘Support vector machine (SVM), as a novel approach in pattern recognition, has demonstrated a success in face detection and face recognition. In this paper, a face recognition approach based on the SVM classifier with the nearest neighbor classifier (NNC) is proposed. The principal component analysis (PCA) is used to reduce the dimension and extract features. Then one-against-all stratedy is used to train the SVM classifiers. At the testing stage, we propose an al-
基金supported by National Natural Science Foundation of China (Nos. 61073133, 60973067, and 61175053)Fundamental Research Funds for the Central Universities of China(No. 2011ZD010)
文摘Numerous models have been proposed to reduce the classification error of Naive Bayes by weakening its attribute independence assumption and some have demonstrated remarkable error performance. Considering that ensemble learning is an effective method of reducing the classifmation error of the classifier, this paper proposes a double-layer Bayesian classifier ensembles （DLBCE） algorithm based on frequent itemsets. DLBCE constructs a double-layer Bayesian classifier （DLBC） for each frequent itemset the new instance contained and finally ensembles all the classifiers by assigning different weight to different classifier according to the conditional mutual information. The experimental results show that the proposed algorithm outperforms other outstanding algorithms.
基金Supported by the National Natural Science Foundation of China (70873117)
文摘This paper proposed an algorithm in which the maximum probability and the weighted average strategy were used for the combination of member classifiers. Using parallel computing, we test the algorithm on a China-Brazil Earth Resources Satellite （CBERS） image for land cover classification. The results show that using three computers in parallel can reduce the classification time by 30%, as compared with using only one computer with a dual core processor. The accuracy of the final image is 93.34%, and Kappa is 0.92. Multiple classifier combination can enhance the precision of the image classification, and parallel computing can increase the speed of calculation so that it becomes possible to process remote sensing images with high efficiency and accuracy.
基金The authors are grateful to MANF UGC,Government of India,for providing financial support under MANF-UGC(MANF-2015-17-JAM-60,506)programme to carry out this work.
文摘The participation of ordinary devices in networking has created a world of connected devices rapidly.The Internet of Things(IoT)includes heterogeneous devices from every field.There are no definite protocols or standards for IoT communication,and most of the IoT devices have limited resources.Enabling a complete security measure for such devices is a challenging task,yet necessary.Many lightweight security solutions have surfaced lately for IoT.The lightweight security protocols are unable to provide an optimum protection against prevailing powerful threats in cyber world.It is also hard to deploy any traditional security protocol on resource-constrained IoT devices.Software-defined networking introduces a centralized control in computer networks.SDN has a programmable approach towards networking that decouples control and data planes.An SDN-based intrusion detection system is proposed which uses deep learning classifier for detection of anomalies in IoT.The proposed intrusion detection system does not burden the IoT devices with security profiles.The proposed work is executed on the simulated environment.The results of the simulation test are evaluated using various matrices and compared with other relevant methods.
文摘Mapping and monitoring the distribution of croplands and crop types support policymakers and international organizations by reducing the risks to food security,notably from climate change and,for that purpose,remote sensing is routinely used.However,identifying specific crop types,cropland,and cropping patterns using space-based observations is challenging because different crop types and cropping patterns have similarity spectral signatures.This study applied a methodology to identify cropland and specific crop types,including tobacco,wheat,barley,and gram,as well as the following cropping patterns:wheat-tobacco,wheat-gram,wheat-barley,and wheat-maize,which are common in Gujranwala District,Pakistan,the study region.The methodology consists of combining optical remote sensing images from Sentinel-2 and Landsat-8 with Machine Learning(ML)methods,namely a Decision Tree Classifier(DTC)and a Random Forest(RF)algorithm.The best time-periods for differentiating cropland from other land cover types were identified,and then Sentinel-2 and Landsat 8 NDVI-based time-series were linked to phenological parameters to determine the different crop types and cropping patterns over the study region using their temporal indices and ML algorithms.The methodology was subsequently evaluated using Landsat images,crop statistical data for 2020 and 2021,and field data on cropping patterns.The results highlight the high level of accuracy of the methodological approach presented using Sentinel-2 and Landsat-8 images,together with ML techniques,for mapping not only the distribution of cropland,but also crop types and cropping patterns when validated at the county level.These results reveal that this methodology has benefits for monitoring and evaluating food security in Pakistan,adding to the evidence base of other studies on the use of remote sensing to identify crop types and cropping patterns in other countries.
基金supported by the National Natural Science Foundation of China(No.61772401)the Fundamental Research Funds for the Central Universities(No.RW180177)supported by the Science and Technology on Communication Information Security Control Laboratory。
文摘Deep learning has been fully verified and accepted in the field of electromagnetic signal classification. However, in many specific scenarios, such as radio resource management for aircraft communications, labeled data are difficult to obtain, which makes the best deep learning methods at present seem almost powerless, because these methods need a large amount of labeled data for training. When the training dataset is small, it is highly possible to fall into overfitting, which causes performance degradation of the deep neural network. For few-shot electromagnetic signal classification, data augmentation is one of the most intuitive countermeasures. In this work, a generative adversarial network based on the data augmentation method is proposed to achieve better classification performance for electromagnetic signals. Based on the similarity principle, a screening mechanism is established to obtain high-quality generated signals. Then, a data union augmentation algorithm is designed by introducing spatiotemporally flipped shapes of the signal. To verify the effectiveness of the proposed data augmentation algorithm, experiments are conducted on the RADIOML 2016.04C dataset and real-world ACARS dataset. The experimental results show that the proposed method significantly improves the performance of few-shot electromagnetic signal classification.
基金supported by National Natural Science Foundation of China (Grant No. 50474035)
文摘The turbo air classifier is widely used powder classification equipment in a variety of fields. The flow field characteristics of the turbo air classifier are important basis for the improvement of the turbo air classifier＇s structural design. The flow field characteristics of the rotor cage in turbo air classifiers were investigated trader different operating conditions by laser Doppler velocimeter（LDV）, and a measure diminishing the axial velocity is proposed. The investigation results show that the tangential velocity of the air flow inside the rotor cage is different from the rotary speed of the rotor cage on the same measurement point due to the influences of both the negative pressure at the exit and the rotation of the rotor cage. The tangential velocity of the air flow likewise decreases as the radius decreases in the case of the rotor cage＇s low rotary speed. In contrast, the tangential velocity of the air flow increases as the radius decreases in the case of the rotor cage＇s high rotary speed. Meanwhile, the vortex inside the rotor cage is found to occur near the pressure side of the blade when the rotor cage＇s rotary speed is less than the tangential velocity of air flow. On the contrary, the vortex is found to occur near the blade suction side once the rotor cage＇s rotary speed is higher than the tangential velocity of air flow. Inside the rotor cage, the axial velocity could not be disregarded and is largely determined by the distances between the measurement point and the exit.
基金Project(2015SK20792)supported by Key Province Key Technology Research and Development Program of the Ministry of Science and Technology of Hunan,ChinaProjects(2019zzts703,2020zzts740,2020zzts202)supported by the Fundamental Research Funds for the Central Universities of ChinaProject(2020P4FZG03A)supported by State Key Laboratory of Vanadium and Titanium Resources Comprehensive Utilization,China。
文摘In this work,the reflux classifier with closely spaced inclined channels is used as the pre-concentration facility to improve the separation efficiency before the shaking table separation.Three operating parameters of reflux classifier(RC)to pre-concentrate fine(0.023−0.15 mm)tailings of antimony oxide were optimized by response surface methodology(RSM)using a three-level Box-Behnken design(BBD).The parameters studied for the optimization were feeding speed,underflow,and ascending water speed.Second-order response functions were produced for the Sb grade and recovery rate of the concentrate.Taking advantage of the quadratic programming,when the factors of feeding,underflow and ascending water are respectively 225,30 and 133 cm^3/min,a better result can be achieved for the concentrate grade of 2.31% and recovery rate of 83.17%.At the same time,70.48% of the tailings with the grade of 0.20% were discarded out of the feeding.The results indicated that the reflux classifier has a good performance in dealing with fine tailings of antimony oxide.Moreover,second-order polynomial equations,ANOVA,and three-dimensional surface plots were developed to evaluate the effects of each parameter on Sb grade and recovery rate of the concentrate.
文摘Predicting stock price movements is a challenging task for academicians and practitioners. In particular, forecasting price movements in emerging markets seems to be more elusive because they are usually more volatile often accompa-nied by thin trading-volumes and they are susceptible to more manipulation compared to mature markets. Technical analysis of stocks and commodities has become a science on its own;quantitative methods and techniques have been applied by many practitioners to forecast price movements. Lagging and sometimes leading technical indicators pro-vide rich quantitative tools for traders and investors in their attempt to gain advantage when making investment or trading decisions. Artificial Neural Networks (ANN) have been used widely in predicting stock prices because of their capability in capturing the non-linearity that often exists in price movements. Recently, Polynomial Classifiers (PC) have been applied to various recognition and classification application and showed favorable results in terms of recog-nition rates and computational complexity as compared to ANN. In this paper, we present two prediction models for predicting securities’ prices. The first model was developed using back propagation feed forward neural networks. The second model was developed using polynomial classifiers (PC), as a first time application for PC to be used in stock prices prediction. The inputs to both models were identical, and both models were trained and tested on the same data. The study was conducted on Dubai Financial Market as an emerging market and applied to two of the market’s leading stocks. In general, both models achieved very good results in terms of mean absolute error percentage. Both models show an average error around 1.5% predicting the next day price, an average error of 2.5% when predicting second day price, and an average error of 4% when predicted the third day price.
