Now object detection based on deep learning tries different strategies.It uses fewer data training networks to achieve the effect of large dataset training.However,the existing methods usually do not achieve the balan...Now object detection based on deep learning tries different strategies.It uses fewer data training networks to achieve the effect of large dataset training.However,the existing methods usually do not achieve the balance between network parameters and training data.It makes the information provided by a small amount of picture data insufficient to optimize model parameters,resulting in unsatisfactory detection results.To improve the accuracy of few shot object detection,this paper proposes a network based on the transformer and high-resolution feature extraction(THR).High-resolution feature extractionmaintains the resolution representation of the image.Channels and spatial attention are used to make the network focus on features that are more useful to the object.In addition,the recently popular transformer is used to fuse the features of the existing object.This compensates for the previous network failure by making full use of existing object features.Experiments on the Pascal VOC and MS-COCO datasets prove that the THR network has achieved better results than previous mainstream few shot object detection.展开更多
As a distributed database,the system security of the blockchain is of great significance to prevent tampering,protect privacy,prevent double spending,and improve credibility.Due to the decentralized and trustless natu...As a distributed database,the system security of the blockchain is of great significance to prevent tampering,protect privacy,prevent double spending,and improve credibility.Due to the decentralized and trustless nature of blockchain,the security defense of the blockchain system has become one of the most important measures.This paper comprehensively reviews the research progress of blockchain security threats and collaborative defense,and we first introduce the overview,classification,and threat assessment process of blockchain security threats.Then,we investigate the research status of single-node defense technology and multi-node collaborative defense technology and summarize the blockchain security evaluation indicators and evaluation methods.Finally,we discuss the challenges of blockchain security and future research directions,such as parallel detection and federated learning.This paper aims to stimulate further research and discussion on blockchain security,providing more reliable security guarantees for the use and development of blockchain technology to face changing threats and challenges through continuous updating and improvement of defense technologies.展开更多
Data prediction can improve the science of decision-making by making predictions about what happens in daily life based on natural law trends.Back propagation(BP)neural network is a widely used prediction method.To re...Data prediction can improve the science of decision-making by making predictions about what happens in daily life based on natural law trends.Back propagation(BP)neural network is a widely used prediction method.To reduce its probability of falling into local optimum and improve the prediction accuracy,we propose an improved BP neural network prediction method based on a multi-strategy sparrow search algorithm(MSSA).The weights and thresholds of the BP neural network are optimized using the sparrow search algorithm(SSA).Three strategies are designed to improve the SSA to enhance its optimization-seeking ability,leading to the MSSA-BP prediction model.The MSSA algorithm was tested with nine different types of benchmark functions to verify the optimization performance of the algorithm.Two different datasets were selected for comparison experiments on three groups of models.Under the same conditions,the mean absolute error(MAE),root mean square error(RMSE),andmean absolute percentage error(MAPE)of the prediction results of MSSA-BPwere significantly reduced,and the convergence speed was significantly improved.MSSA-BP can effectively improve the prediction accuracy and has certain application value.展开更多
With the increasing number of smart devices and the development of machine learning technology,the value of users’personal data is becoming more and more important.Based on the premise of protecting users’personal p...With the increasing number of smart devices and the development of machine learning technology,the value of users’personal data is becoming more and more important.Based on the premise of protecting users’personal privacy data,federated learning(FL)uses data stored on edge devices to realize training tasks by contributing training model parameters without revealing the original data.However,since FL can still leak the user’s original data by exchanging gradient information.The existing privacy protection strategy will increase the uplink time due to encryption measures.It is a huge challenge in terms of communication.When there are a large number of devices,the privacy protection cost of the system is higher.Based on these issues,we propose a privacy-preserving scheme of user-based group collaborative federated learning(GrCol-PPFL).Our scheme primarily divides participants into several groups and each group communicates in a chained transmission mechanism.All groups work in parallel at the same time.The server distributes a random parameter with the same dimension as the model parameter for each participant as a mask for the model parameter.We use the public datasets of modified national institute of standards and technology database(MNIST)to test the model accuracy.The experimental results show that GrCol-PPFL not only ensures the accuracy of themodel,but also ensures the security of the user’s original data when users collude with each other.Finally,through numerical experiments,we show that by changing the number of groups,we can find the optimal number of groups that reduces the uplink consumption time.展开更多
With the rapid development of blockchain technology,the number of distributed applications continues to increase,so ensuring the security of the network has become particularly important.However,due to its decentraliz...With the rapid development of blockchain technology,the number of distributed applications continues to increase,so ensuring the security of the network has become particularly important.However,due to its decentralized,decentralized nature,blockchain networks are vulnerable to distributed denial-of-service(DDoS)attacks,which can lead to service stops,causing serious economic losses and social impacts.The research questions in this paper mainly include two aspects:first,the classification of DDoS,which refers to detecting whether blockchain nodes are suffering DDoS attacks,that is,detecting the data of nodes in parallel;The second is the problem of DDoS segmentation,that is,multiple pieces of data that appear at the same time are determined which type of DDoS attack they belong to.In order to solve these problems,this paper proposes a modified PointNet(MPointNet)for the classification and type segmentation of DDoS attacks.A dataset containing multiple DDoS attack types was constructed using the CIC-DDoS2019 dataset,and trained,validated,and tested accordingly.The results show that the proposed DDoS attack classification method has high performance and can be used for the actual blockchain security maintenance process.The accuracy rate of classification tasks reached 99.65%,and the accuracy of type segmentation tasks reached 85.47%.Therefore,the method proposed in this paper has high application value in detecting the classification and segmentation of DDoS attacks.展开更多
Interpreting deep neural networks is of great importance to understand and verify deep models for natural language processing(NLP)tasks.However,most existing approaches only focus on improving the performance of model...Interpreting deep neural networks is of great importance to understand and verify deep models for natural language processing(NLP)tasks.However,most existing approaches only focus on improving the performance of models but ignore their interpretability.In this work,we propose a Randomly Wired Graph Neural Network(RWGNN)by using graph to model the structure of Neural Network,which could solve two major problems(word-boundary ambiguity and polysemy)of ChineseNER.Besides,we develop a pipeline to explain the RWGNNby using Saliency Map and Adversarial Attacks.Experimental results demonstrate that our approach can identify meaningful and reasonable interpretations for hidden states of RWGNN.展开更多
Image semantic segmentation is an important branch of computer vision of a wide variety of practical applications such as medical image analysis,autonomous driving,virtual or augmented reality,etc.In recent years,due ...Image semantic segmentation is an important branch of computer vision of a wide variety of practical applications such as medical image analysis,autonomous driving,virtual or augmented reality,etc.In recent years,due to the remarkable performance of transformer and multilayer perceptron(MLP)in computer vision,which is equivalent to convolutional neural network(CNN),there has been a substantial amount of image semantic segmentation works aimed at developing different types of deep learning architecture.This survey aims to provide a comprehensive overview of deep learning methods in the field of general image semantic segmentation.Firstly,the commonly used image segmentation datasets are listed.Next,extensive pioneering works are deeply studied from multiple perspectives(e.g.,network structures,feature fusion methods,attention mechanisms),and are divided into four categories according to different network architectures:CNN-based architectures,transformer-based architectures,MLP-based architectures,and others.Furthermore,this paper presents some common evaluation metrics and compares the respective advantages and limitations of popular techniques both in terms of architectural design and their experimental value on the most widely used datasets.Finally,possible future research directions and challenges are discussed for the reference of other researchers.展开更多
In the realm of Multi-Label Text Classification(MLTC),the dual challenges of extracting rich semantic features from text and discerning inter-label relationships have spurred innovative approaches.Many studies in sema...In the realm of Multi-Label Text Classification(MLTC),the dual challenges of extracting rich semantic features from text and discerning inter-label relationships have spurred innovative approaches.Many studies in semantic feature extraction have turned to external knowledge to augment the model’s grasp of textual content,often overlooking intrinsic textual cues such as label statistical features.In contrast,these endogenous insights naturally align with the classification task.In our paper,to complement this focus on intrinsic knowledge,we introduce a novel Gate-Attention mechanism.This mechanism adeptly integrates statistical features from the text itself into the semantic fabric,enhancing the model’s capacity to understand and represent the data.Additionally,to address the intricate task of mining label correlations,we propose a Dual-end enhancement mechanism.This mechanism effectively mitigates the challenges of information loss and erroneous transmission inherent in traditional long short term memory propagation.We conducted an extensive battery of experiments on the AAPD and RCV1-2 datasets.These experiments serve the dual purpose of confirming the efficacy of both the Gate-Attention mechanism and the Dual-end enhancement mechanism.Our final model unequivocally outperforms the baseline model,attesting to its robustness.These findings emphatically underscore the imperativeness of taking into account not just external knowledge but also the inherent intricacies of textual data when crafting potent MLTC models.展开更多
Distributed denial-of-service(DDoS)is a rapidly growing problem with the fast development of the Internet.There are multitude DDoS detection approaches,however,three major problems about DDoS attack detection appear i...Distributed denial-of-service(DDoS)is a rapidly growing problem with the fast development of the Internet.There are multitude DDoS detection approaches,however,three major problems about DDoS attack detection appear in the big data environment.Firstly,to shorten the respond time of the DDoS attack detector;secondly,to reduce the required compute resources;lastly,to achieve a high detection rate with low false alarm rate.In the paper,we propose an abnormal network flow feature sequence prediction approach which could fit to be used as a DDoS attack detector in the big data environment and solve aforementioned problems.We define a network flow abnormal index as PDRA with the percentage of old IP addresses,the increment of the new IP addresses,the ratio of new IP addresses to the old IP addresses and average accessing rate of each new IP address.We design an IP address database using sequential storage model which has a constant time complexity.The autoregressive integrated moving average(ARIMA)trending prediction module will be started if and only if the number of continuous PDRA sequence value,which all exceed an PDRA abnormal threshold(PAT),reaches a certain preset threshold.And then calculate the probability that is the percentage of forecasting PDRA sequence value which exceed the PAT.Finally we identify the DDoS attack based on the abnormal probability of the forecasting PDRA sequence.Both theorem and experiment show that the method we proposed can effectively reduce the compute resources consumption,identify DDoS attack at its initial stage with higher detection rate and lower false alarm rate.展开更多
Image recognition has always been a hot research topic in the scientific community and industry.The emergence of convolutional neural networks(CNN)has made this technology turned into research focus on the field of co...Image recognition has always been a hot research topic in the scientific community and industry.The emergence of convolutional neural networks(CNN)has made this technology turned into research focus on the field of computer vision,especially in image recognition.But it makes the recognition result largely dependent on the number and quality of training samples.Recently,DCGAN has become a frontier method for generating images,sounds,and videos.In this paper,DCGAN is used to generate sample that is difficult to collect and proposed an efficient design method of generating model.We combine DCGAN with CNN for the second time.Use DCGAN to generate samples and training in image recognition model,which based by CNN.This method can enhance the classification model and effectively improve the accuracy of image recognition.In the experiment,we used the radar profile as dataset for 4 categories and achieved satisfactory classification performance.This paper applies image recognition technology to the meteorological field.展开更多
Accurate segmentation of CT images of liver tumors is an important adjunct for the liver diagnosis and treatment of liver diseases.In recent years,due to the great improvement of hard device,many deep learning based m...Accurate segmentation of CT images of liver tumors is an important adjunct for the liver diagnosis and treatment of liver diseases.In recent years,due to the great improvement of hard device,many deep learning based methods have been proposed for automatic liver segmentation.Among them,there are the plain neural network headed by FCN and the residual neural network headed by Resnet,both of which have many variations.They have achieved certain achievements in medical image segmentation.In this paper,we firstly select five representative structures,i.e.,FCN,U-Net,Segnet,Resnet and Densenet,to investigate their performance on liver segmentation.Since original Resnet and Densenet could not perform image segmentation directly,we make some adjustments for them to perform live segmentation.Our experimental results show that Densenet performs the best on liver segmentation,followed by Resnet.Both perform much better than Segnet,U-Net,and FCN.Among Segnet,U-Net,and FCN,U-Net performs the best,followed by Segnet.FCN performs the worst.展开更多
The accuracy of predicting the Producer Price Index(PPI)plays an indispensable role in government economic work.However,it is difficult to forecast the PPI.In our research,we first propose an unprecedented hybrid mode...The accuracy of predicting the Producer Price Index(PPI)plays an indispensable role in government economic work.However,it is difficult to forecast the PPI.In our research,we first propose an unprecedented hybrid model based on fuzzy information granulation that integrates the GA-SVR and ARIMA(Autoregressive Integrated Moving Average Model)models.The fuzzy-information-granulation-based GA-SVR-ARIMA hybrid model is intended to deal with the problem of imprecision in PPI estimation.The proposed model adopts the fuzzy information-granulation algorithm to pre-classification-process monthly training samples of the PPI,and produced three different sequences of fuzzy information granules,whose Support Vector Regression(SVR)machine forecast models were separately established for their Genetic Algorithm(GA)optimization parameters.Finally,the residual errors of the GA-SVR model were rectified through ARIMA modeling,and the PPI estimate was reached.Research shows that the PPI value predicted by this hybrid model is more accurate than that predicted by other models,including ARIMA,GRNN,and GA-SVR,following several comparative experiments.Research also indicates the precision and validation of the PPI prediction of the hybrid model and demonstrates that the model has consistent ability to leverage the forecasting advantage of GA-SVR in non-linear space and of ARIMA in linear space.展开更多
The research on the governing blockchain by blockchain supervision system is an important development trend of blockchain technology.In this system there is a supervisory blockchain managing and governing the supervis...The research on the governing blockchain by blockchain supervision system is an important development trend of blockchain technology.In this system there is a supervisory blockchain managing and governing the supervised blockchain based on blockchain technology,results in a uniquely cross-blockchain demand to consensus mechanism for solving the trust problem between supervisory blockchain and supervised blockchain.To solve this problem,this paper proposes a cross-blockchain consensus mechanism based on smart contract and a set of smart contracts endorse the crossblockchain consensus.New consensus mechanism called Proof-of-EndorseContracts(PoEC)consensus,which firstly transfers the consensus reached in supervisory blockchain to supervised blockchain by supervisory nodes,then packages the supervisory block in supervisory blockchain and transmits it to the smart contract deployed in the supervised blockchain,finally miners in supervised blockchain will execute and package the new block according to the status of the smart contract.The core part of the consensus mechanism is Endorse Contracts which designed and implemented by us and verified the effectiveness through experiments.PoEC consensus mechanism and Endorse Contracts support the supervised blockchain to join the governing blockchain by blockchain system without changing the original consensus mechanism,which has the advantages of low cost,high scalability and being able to crossblockchain.This paper proves that our method can provide a feasible crossblockchain governance scheme for the field of blockchain governance.展开更多
Insects play an important role in the natural ecology,it is of great significance for ecology to research on insects.Nowadays,the invasion of alien species has brought serious troubles and a lot of losses to local lif...Insects play an important role in the natural ecology,it is of great significance for ecology to research on insects.Nowadays,the invasion of alien species has brought serious troubles and a lot of losses to local life.However,there is still much room for improvement in the accuracy of insect recognition to effectively prevent the invasion of alien species.As the latest target detection algorithm,YOLOv5 has been used in various scene detection tasks,because of its powerful recognition capabilities and extremely high accuracy.As the problem of imbalance of feature maps at different scales will affect the accuracy of recognition,we propose that adding an attention mechanism based on YOLOv5.The channel attention module and the spatial attention module are added to highlight the important information in the feature map and weaken the secondary information,enhancing the recognition ability of the network.Through training on self-made insect data sets,experimental results show that the mAP@0.5 value reaches 92.5%and the F1 score reaches 0.91.Compared with YOLOv5,the map has increased by 1.7%,and the F1 score has increased by 0.02,proving the effectiveness of insect recognition based on improved YOLOv5.In conclusion,we provide effective technical support for insect identification,especially for pest identification.展开更多
Precipitation nowcasting is of great significance for severe convective weather warnings.Radar echo extrapolation is a commonly used precipitation nowcasting method.However,the traditional radar echo extrapolation met...Precipitation nowcasting is of great significance for severe convective weather warnings.Radar echo extrapolation is a commonly used precipitation nowcasting method.However,the traditional radar echo extrapolation methods are encountered with the dilemma of low prediction accuracy and extrapolation ambiguity.The reason is that those methods cannot retain important long-term information and fail to capture short-term motion information from the long-range data stream.In order to solve the above problems,we select the spatiotemporal long short-term memory(ST-LSTM)as the recurrent unit of the model and integrate the 3D convolution operation in it to strengthen the model’s ability to capture short-term motion information which plays a vital role in the prediction of radar echo motion trends.For the purpose of enhancing the model’s ability to retain long-term important information,we also introduce the channel attention mechanism to achieve this goal.In the experiment,the training and testing datasets are constructed using radar data of Shanghai,we compare our model with three benchmark models under the reflectance thresholds of 15 and 25.Experimental results demonstrate that the proposed model outperforms the three benchmark models in radar echo extrapolation task,which obtains a higher accuracy rate and improves the clarity of the extrapolated image.展开更多
PM2.5 concentration prediction is of great significance to environmental protection and human health.Achieving accurate prediction of PM2.5 concentration has become an important research task.However,PM2.5 pollutants ...PM2.5 concentration prediction is of great significance to environmental protection and human health.Achieving accurate prediction of PM2.5 concentration has become an important research task.However,PM2.5 pollutants can spread in the earth’s atmosphere,causing mutual influence between different cities.To effectively capture the air pollution relationship between cities,this paper proposes a novel spatiotemporal model combining graph attention neural network(GAT)and gated recurrent unit(GRU),named GAT-GRU for PM2.5 concentration prediction.Specifically,GAT is used to learn the spatial dependence of PM2.5 concentration data in different cities,and GRU is to extract the temporal dependence of the long-term data series.The proposed model integrates the learned spatio-temporal dependencies to capture long-term complex spatio-temporal features.Considering that air pollution is related to the meteorological conditions of the city,the knowledge acquired from meteorological data is used in the model to enhance PM2.5 prediction performance.The input of the GAT-GRU model consists of PM2.5 concentration data and meteorological data.In order to verify the effectiveness of the proposed GAT-GRU prediction model,this paper designs experiments on real-world datasets compared with other baselines.Experimental results prove that our model achieves excellent performance in PM2.5 concentration prediction.展开更多
Image inpainting based on deep learning has been greatly improved.The original purpose of image inpainting was to repair some broken photos, suchas inpainting artifacts. However, it may also be used for malicious oper...Image inpainting based on deep learning has been greatly improved.The original purpose of image inpainting was to repair some broken photos, suchas inpainting artifacts. However, it may also be used for malicious operations,such as destroying evidence. Therefore, detection and localization of imageinpainting operations are essential. Recent research shows that high-pass filteringfull convolutional network (HPFCN) is applied to image inpainting detection andachieves good results. However, those methods did not consider the spatial location and channel information of the feature map. To solve these shortcomings, weintroduce the squeezed excitation blocks (SE) and propose a high-pass filter attention full convolutional network (HPACN). In feature extraction, we apply concurrent spatial and channel attention (scSE) to enhance feature extraction and obtainmore information. Channel attention (cSE) is introduced in upsampling toenhance detection and localization. The experimental results show that the proposed method can achieve improvement on ImageNet.展开更多
Multi-target regression is concerned with the simultaneous prediction of multiple continuous target variables based on the same set of input variables.It has received relatively small attention from the Machine Learni...Multi-target regression is concerned with the simultaneous prediction of multiple continuous target variables based on the same set of input variables.It has received relatively small attention from the Machine Learning community.However,multi-target regression exists in many real-world applications.In this paper we conduct extensive experiments to investigate the performance of three representative multi-target regression learning algorithms(i.e.Multi-Target Stacking(MTS),Random Linear Target Combination(RLTC),and Multi-Objective Random Forest(MORF)),comparing the baseline single-target learning.Our experimental results show that all three multi-target regression learning algorithms do improve the performance of the single-target learning.Among them,MTS performs the best,followed by RLTC,followed by MORF.However,the single-target learning sometimes still performs very well,even the best.This analysis sheds the light on multi-target regression learning and indicates that the single-target learning is a competitive baseline for multi-target regression learning on multi-target domains.展开更多
The pattern of thematic progression,reflecting the semantic relationships between contextual two sentences,is an important subject in discourse analysis.We introduce a new corpus of Chinese news discourses annotated w...The pattern of thematic progression,reflecting the semantic relationships between contextual two sentences,is an important subject in discourse analysis.We introduce a new corpus of Chinese news discourses annotated with thematic progression information and explore some computational methods to automatically extracting the discourse structural features of simplified thematic progression pattern(STPP)between contextual sentences in a text.Furthermore,these features are used in a hybrid approach to a major discourse analysis task,Chinese coreference resolution.This novel approach is built up via heuristic sieves and a machine learning method that comprehensively utilizes both the top-down STPP features and the bottom-up semantic features.Experimental results on the intersection of the CoNLL-2012 task shared dataset and the CDTC corpus demonstrate the effectiveness of our proposed approach.展开更多
基金the National Natural Science Foundation of China under grant 62172059 and 62072055Hunan Provincial Natural Science Foundations of China under Grant 2020JJ4626+2 种基金Scientific Research Fund of Hunan Provincial Education Department of China under Grant 19B004“Double First-class”International Cooperation and Development Scientific Research Project of Changsha University of Science and Technology under Grant 2018IC25the Young Teacher Growth Plan Project of Changsha University of Science and Technology under Grant 2019QJCZ076.
文摘Now object detection based on deep learning tries different strategies.It uses fewer data training networks to achieve the effect of large dataset training.However,the existing methods usually do not achieve the balance between network parameters and training data.It makes the information provided by a small amount of picture data insufficient to optimize model parameters,resulting in unsatisfactory detection results.To improve the accuracy of few shot object detection,this paper proposes a network based on the transformer and high-resolution feature extraction(THR).High-resolution feature extractionmaintains the resolution representation of the image.Channels and spatial attention are used to make the network focus on features that are more useful to the object.In addition,the recently popular transformer is used to fuse the features of the existing object.This compensates for the previous network failure by making full use of existing object features.Experiments on the Pascal VOC and MS-COCO datasets prove that the THR network has achieved better results than previous mainstream few shot object detection.
基金supported by National Natural Science Foundation of China(Grant Nos.62162022 and 62162024)Young Talents’Science and Technology Innovation Project of Hainan Association for Science and Technology(Grant No.QCXM202007)Hainan Provincial Natural Science Foundation of China(Grant Nos.2019RC098 and 621RC612).
文摘As a distributed database,the system security of the blockchain is of great significance to prevent tampering,protect privacy,prevent double spending,and improve credibility.Due to the decentralized and trustless nature of blockchain,the security defense of the blockchain system has become one of the most important measures.This paper comprehensively reviews the research progress of blockchain security threats and collaborative defense,and we first introduce the overview,classification,and threat assessment process of blockchain security threats.Then,we investigate the research status of single-node defense technology and multi-node collaborative defense technology and summarize the blockchain security evaluation indicators and evaluation methods.Finally,we discuss the challenges of blockchain security and future research directions,such as parallel detection and federated learning.This paper aims to stimulate further research and discussion on blockchain security,providing more reliable security guarantees for the use and development of blockchain technology to face changing threats and challenges through continuous updating and improvement of defense technologies.
基金the National Natural Science Foundation of China(Grant No.62162024 and 62162022)Key Projects in Hainan Province(Grant ZDYF2021GXJS003 and Grant ZDYF2020040)the Major science and technology project of Hainan Province(Grant No.ZDKJ2020012).
文摘Data prediction can improve the science of decision-making by making predictions about what happens in daily life based on natural law trends.Back propagation(BP)neural network is a widely used prediction method.To reduce its probability of falling into local optimum and improve the prediction accuracy,we propose an improved BP neural network prediction method based on a multi-strategy sparrow search algorithm(MSSA).The weights and thresholds of the BP neural network are optimized using the sparrow search algorithm(SSA).Three strategies are designed to improve the SSA to enhance its optimization-seeking ability,leading to the MSSA-BP prediction model.The MSSA algorithm was tested with nine different types of benchmark functions to verify the optimization performance of the algorithm.Two different datasets were selected for comparison experiments on three groups of models.Under the same conditions,the mean absolute error(MAE),root mean square error(RMSE),andmean absolute percentage error(MAPE)of the prediction results of MSSA-BPwere significantly reduced,and the convergence speed was significantly improved.MSSA-BP can effectively improve the prediction accuracy and has certain application value.
基金supported by the Major science and technology project of Hainan Province(Grant No.ZDKJ2020012)National Natural Science Foundation of China(Grant No.62162024 and 62162022)Key Projects in Hainan Province(Grant ZDYF2021GXJS003 and Grant ZDYF2020040).
文摘With the increasing number of smart devices and the development of machine learning technology,the value of users’personal data is becoming more and more important.Based on the premise of protecting users’personal privacy data,federated learning(FL)uses data stored on edge devices to realize training tasks by contributing training model parameters without revealing the original data.However,since FL can still leak the user’s original data by exchanging gradient information.The existing privacy protection strategy will increase the uplink time due to encryption measures.It is a huge challenge in terms of communication.When there are a large number of devices,the privacy protection cost of the system is higher.Based on these issues,we propose a privacy-preserving scheme of user-based group collaborative federated learning(GrCol-PPFL).Our scheme primarily divides participants into several groups and each group communicates in a chained transmission mechanism.All groups work in parallel at the same time.The server distributes a random parameter with the same dimension as the model parameter for each participant as a mask for the model parameter.We use the public datasets of modified national institute of standards and technology database(MNIST)to test the model accuracy.The experimental results show that GrCol-PPFL not only ensures the accuracy of themodel,but also ensures the security of the user’s original data when users collude with each other.Finally,through numerical experiments,we show that by changing the number of groups,we can find the optimal number of groups that reduces the uplink consumption time.
基金This work was supported by Hainan Provincial Natural Science Foundation of China(Grant No.2019RC098,Grant No.723QN238 and Grant No.621RC612)National Natural Science Foundation of China(Grant No.62162022 and 62162024)+1 种基金Key Projects in Hainan Province(GrantZDYF2020040 andGrantZDYF2020033)Young Talents’Science and Technology Innovation Project of Hainan Association for Science and Technology(Grant No.QCXM202007).
文摘With the rapid development of blockchain technology,the number of distributed applications continues to increase,so ensuring the security of the network has become particularly important.However,due to its decentralized,decentralized nature,blockchain networks are vulnerable to distributed denial-of-service(DDoS)attacks,which can lead to service stops,causing serious economic losses and social impacts.The research questions in this paper mainly include two aspects:first,the classification of DDoS,which refers to detecting whether blockchain nodes are suffering DDoS attacks,that is,detecting the data of nodes in parallel;The second is the problem of DDoS segmentation,that is,multiple pieces of data that appear at the same time are determined which type of DDoS attack they belong to.In order to solve these problems,this paper proposes a modified PointNet(MPointNet)for the classification and type segmentation of DDoS attacks.A dataset containing multiple DDoS attack types was constructed using the CIC-DDoS2019 dataset,and trained,validated,and tested accordingly.The results show that the proposed DDoS attack classification method has high performance and can be used for the actual blockchain security maintenance process.The accuracy rate of classification tasks reached 99.65%,and the accuracy of type segmentation tasks reached 85.47%.Therefore,the method proposed in this paper has high application value in detecting the classification and segmentation of DDoS attacks.
基金supported by the National Science Foundation of China(NSFC)underGrants 61876217 and 62176175the Innovative Team of Jiangsu Province under Grant XYDXX-086Jiangsu Postgraduate Research and Innovation Plan(KYCX20_2762).
文摘Interpreting deep neural networks is of great importance to understand and verify deep models for natural language processing(NLP)tasks.However,most existing approaches only focus on improving the performance of models but ignore their interpretability.In this work,we propose a Randomly Wired Graph Neural Network(RWGNN)by using graph to model the structure of Neural Network,which could solve two major problems(word-boundary ambiguity and polysemy)of ChineseNER.Besides,we develop a pipeline to explain the RWGNNby using Saliency Map and Adversarial Attacks.Experimental results demonstrate that our approach can identify meaningful and reasonable interpretations for hidden states of RWGNN.
基金supported by the Major science and technology project of Hainan Province(Grant No.ZDKJ2020012)National Natural Science Foundation of China(Grant No.62162024 and 62162022)+1 种基金Key Projects in Hainan Province(Grant ZDYF2021GXJS003 and Grant ZDYF2020040)Graduate Innovation Project(Grant No.Qhys2021-187).
文摘Image semantic segmentation is an important branch of computer vision of a wide variety of practical applications such as medical image analysis,autonomous driving,virtual or augmented reality,etc.In recent years,due to the remarkable performance of transformer and multilayer perceptron(MLP)in computer vision,which is equivalent to convolutional neural network(CNN),there has been a substantial amount of image semantic segmentation works aimed at developing different types of deep learning architecture.This survey aims to provide a comprehensive overview of deep learning methods in the field of general image semantic segmentation.Firstly,the commonly used image segmentation datasets are listed.Next,extensive pioneering works are deeply studied from multiple perspectives(e.g.,network structures,feature fusion methods,attention mechanisms),and are divided into four categories according to different network architectures:CNN-based architectures,transformer-based architectures,MLP-based architectures,and others.Furthermore,this paper presents some common evaluation metrics and compares the respective advantages and limitations of popular techniques both in terms of architectural design and their experimental value on the most widely used datasets.Finally,possible future research directions and challenges are discussed for the reference of other researchers.
基金supported by National Natural Science Foundation of China(NSFC)(Grant Nos.62162022,62162024)the Key Research and Development Program of Hainan Province(Grant Nos.ZDYF2020040,ZDYF2021GXJS003)+2 种基金the Major Science and Technology Project of Hainan Province(Grant No.ZDKJ2020012)Hainan Provincial Natural Science Foundation of China(Grant Nos.620MS021,621QN211)Science and Technology Development Center of the Ministry of Education Industry-University-Research Innovation Fund(2021JQR017).
文摘In the realm of Multi-Label Text Classification(MLTC),the dual challenges of extracting rich semantic features from text and discerning inter-label relationships have spurred innovative approaches.Many studies in semantic feature extraction have turned to external knowledge to augment the model’s grasp of textual content,often overlooking intrinsic textual cues such as label statistical features.In contrast,these endogenous insights naturally align with the classification task.In our paper,to complement this focus on intrinsic knowledge,we introduce a novel Gate-Attention mechanism.This mechanism adeptly integrates statistical features from the text itself into the semantic fabric,enhancing the model’s capacity to understand and represent the data.Additionally,to address the intricate task of mining label correlations,we propose a Dual-end enhancement mechanism.This mechanism effectively mitigates the challenges of information loss and erroneous transmission inherent in traditional long short term memory propagation.We conducted an extensive battery of experiments on the AAPD and RCV1-2 datasets.These experiments serve the dual purpose of confirming the efficacy of both the Gate-Attention mechanism and the Dual-end enhancement mechanism.Our final model unequivocally outperforms the baseline model,attesting to its robustness.These findings emphatically underscore the imperativeness of taking into account not just external knowledge but also the inherent intricacies of textual data when crafting potent MLTC models.
基金This work was supported by the National Natural Science Foundation of China[No.61762033,61363071,61702539]The National Natural Science Foundation of Hainan[No.617048,2018CXTD333]+1 种基金Hainan University Doctor Start Fund Project[No.kyqd1328]Hainan University Youth Fund Project[No.qnjj1444].
文摘Distributed denial-of-service(DDoS)is a rapidly growing problem with the fast development of the Internet.There are multitude DDoS detection approaches,however,three major problems about DDoS attack detection appear in the big data environment.Firstly,to shorten the respond time of the DDoS attack detector;secondly,to reduce the required compute resources;lastly,to achieve a high detection rate with low false alarm rate.In the paper,we propose an abnormal network flow feature sequence prediction approach which could fit to be used as a DDoS attack detector in the big data environment and solve aforementioned problems.We define a network flow abnormal index as PDRA with the percentage of old IP addresses,the increment of the new IP addresses,the ratio of new IP addresses to the old IP addresses and average accessing rate of each new IP address.We design an IP address database using sequential storage model which has a constant time complexity.The autoregressive integrated moving average(ARIMA)trending prediction module will be started if and only if the number of continuous PDRA sequence value,which all exceed an PDRA abnormal threshold(PAT),reaches a certain preset threshold.And then calculate the probability that is the percentage of forecasting PDRA sequence value which exceed the PAT.Finally we identify the DDoS attack based on the abnormal probability of the forecasting PDRA sequence.Both theorem and experiment show that the method we proposed can effectively reduce the compute resources consumption,identify DDoS attack at its initial stage with higher detection rate and lower false alarm rate.
文摘Image recognition has always been a hot research topic in the scientific community and industry.The emergence of convolutional neural networks(CNN)has made this technology turned into research focus on the field of computer vision,especially in image recognition.But it makes the recognition result largely dependent on the number and quality of training samples.Recently,DCGAN has become a frontier method for generating images,sounds,and videos.In this paper,DCGAN is used to generate sample that is difficult to collect and proposed an efficient design method of generating model.We combine DCGAN with CNN for the second time.Use DCGAN to generate samples and training in image recognition model,which based by CNN.This method can enhance the classification model and effectively improve the accuracy of image recognition.In the experiment,we used the radar profile as dataset for 4 categories and achieved satisfactory classification performance.This paper applies image recognition technology to the meteorological field.
基金This research has been partially supported by National Science Foundation under grant IIS-1115417the National Natural Science Foundation of China under grant 61728205,61876217+1 种基金the“double first-class”international cooperation and development scientific research project of Changsha University of Science and Technology(No.2018IC25)the Science and Technology Development Project of Suzhou under grant SZS201609 and SYG201707.
文摘Accurate segmentation of CT images of liver tumors is an important adjunct for the liver diagnosis and treatment of liver diseases.In recent years,due to the great improvement of hard device,many deep learning based methods have been proposed for automatic liver segmentation.Among them,there are the plain neural network headed by FCN and the residual neural network headed by Resnet,both of which have many variations.They have achieved certain achievements in medical image segmentation.In this paper,we firstly select five representative structures,i.e.,FCN,U-Net,Segnet,Resnet and Densenet,to investigate their performance on liver segmentation.Since original Resnet and Densenet could not perform image segmentation directly,we make some adjustments for them to perform live segmentation.Our experimental results show that Densenet performs the best on liver segmentation,followed by Resnet.Both perform much better than Segnet,U-Net,and FCN.Among Segnet,U-Net,and FCN,U-Net performs the best,followed by Segnet.FCN performs the worst.
基金This work was supported by Hainan Provincial Natural Science Foundation of China[2018CXTD333,617048]The National Natural Science Foundation of China[61762033,61702539]+1 种基金Hainan University Doctor Start Fund Project[kyqd1328]Hainan University Youth Fund Project[qnjj1444].
文摘The accuracy of predicting the Producer Price Index(PPI)plays an indispensable role in government economic work.However,it is difficult to forecast the PPI.In our research,we first propose an unprecedented hybrid model based on fuzzy information granulation that integrates the GA-SVR and ARIMA(Autoregressive Integrated Moving Average Model)models.The fuzzy-information-granulation-based GA-SVR-ARIMA hybrid model is intended to deal with the problem of imprecision in PPI estimation.The proposed model adopts the fuzzy information-granulation algorithm to pre-classification-process monthly training samples of the PPI,and produced three different sequences of fuzzy information granules,whose Support Vector Regression(SVR)machine forecast models were separately established for their Genetic Algorithm(GA)optimization parameters.Finally,the residual errors of the GA-SVR model were rectified through ARIMA modeling,and the PPI estimate was reached.Research shows that the PPI value predicted by this hybrid model is more accurate than that predicted by other models,including ARIMA,GRNN,and GA-SVR,following several comparative experiments.Research also indicates the precision and validation of the PPI prediction of the hybrid model and demonstrates that the model has consistent ability to leverage the forecasting advantage of GA-SVR in non-linear space and of ARIMA in linear space.
基金This work was supported by National Natural Science Foundation of China(Grant No.62162022 and 62162024)Key Projects in Hainan Province(Grant ZDYF2021GXJS003 and Grant ZDYF2020040)the Major science and technology project of Hainan Province(Grant No.ZDKJ2020012).
文摘The research on the governing blockchain by blockchain supervision system is an important development trend of blockchain technology.In this system there is a supervisory blockchain managing and governing the supervised blockchain based on blockchain technology,results in a uniquely cross-blockchain demand to consensus mechanism for solving the trust problem between supervisory blockchain and supervised blockchain.To solve this problem,this paper proposes a cross-blockchain consensus mechanism based on smart contract and a set of smart contracts endorse the crossblockchain consensus.New consensus mechanism called Proof-of-EndorseContracts(PoEC)consensus,which firstly transfers the consensus reached in supervisory blockchain to supervised blockchain by supervisory nodes,then packages the supervisory block in supervisory blockchain and transmits it to the smart contract deployed in the supervised blockchain,finally miners in supervised blockchain will execute and package the new block according to the status of the smart contract.The core part of the consensus mechanism is Endorse Contracts which designed and implemented by us and verified the effectiveness through experiments.PoEC consensus mechanism and Endorse Contracts support the supervised blockchain to join the governing blockchain by blockchain system without changing the original consensus mechanism,which has the advantages of low cost,high scalability and being able to crossblockchain.This paper proves that our method can provide a feasible crossblockchain governance scheme for the field of blockchain governance.
基金This work is supported by the National Natural Science Foundation of China(Grant No.42075007)the Open Grants of the State Key Laboratory of Severe Weather(No.2021LASW-B19).
文摘Insects play an important role in the natural ecology,it is of great significance for ecology to research on insects.Nowadays,the invasion of alien species has brought serious troubles and a lot of losses to local life.However,there is still much room for improvement in the accuracy of insect recognition to effectively prevent the invasion of alien species.As the latest target detection algorithm,YOLOv5 has been used in various scene detection tasks,because of its powerful recognition capabilities and extremely high accuracy.As the problem of imbalance of feature maps at different scales will affect the accuracy of recognition,we propose that adding an attention mechanism based on YOLOv5.The channel attention module and the spatial attention module are added to highlight the important information in the feature map and weaken the secondary information,enhancing the recognition ability of the network.Through training on self-made insect data sets,experimental results show that the mAP@0.5 value reaches 92.5%and the F1 score reaches 0.91.Compared with YOLOv5,the map has increased by 1.7%,and the F1 score has increased by 0.02,proving the effectiveness of insect recognition based on improved YOLOv5.In conclusion,we provide effective technical support for insect identification,especially for pest identification.
基金This work was supported by the National Natural Science Foundation of China(Grant No.42075007)the Open Grants of the State Key Laboratory of Severe Weather(No.2021LASW-B19).
文摘Precipitation nowcasting is of great significance for severe convective weather warnings.Radar echo extrapolation is a commonly used precipitation nowcasting method.However,the traditional radar echo extrapolation methods are encountered with the dilemma of low prediction accuracy and extrapolation ambiguity.The reason is that those methods cannot retain important long-term information and fail to capture short-term motion information from the long-range data stream.In order to solve the above problems,we select the spatiotemporal long short-term memory(ST-LSTM)as the recurrent unit of the model and integrate the 3D convolution operation in it to strengthen the model’s ability to capture short-term motion information which plays a vital role in the prediction of radar echo motion trends.For the purpose of enhancing the model’s ability to retain long-term important information,we also introduce the channel attention mechanism to achieve this goal.In the experiment,the training and testing datasets are constructed using radar data of Shanghai,we compare our model with three benchmark models under the reflectance thresholds of 15 and 25.Experimental results demonstrate that the proposed model outperforms the three benchmark models in radar echo extrapolation task,which obtains a higher accuracy rate and improves the clarity of the extrapolated image.
基金Authors The research project is partially supported by National Natural ScienceFoundation of China under Grant No. 62072015, U19B2039, U1811463National Key R&D Programof China 2018YFB1600903.
文摘PM2.5 concentration prediction is of great significance to environmental protection and human health.Achieving accurate prediction of PM2.5 concentration has become an important research task.However,PM2.5 pollutants can spread in the earth’s atmosphere,causing mutual influence between different cities.To effectively capture the air pollution relationship between cities,this paper proposes a novel spatiotemporal model combining graph attention neural network(GAT)and gated recurrent unit(GRU),named GAT-GRU for PM2.5 concentration prediction.Specifically,GAT is used to learn the spatial dependence of PM2.5 concentration data in different cities,and GRU is to extract the temporal dependence of the long-term data series.The proposed model integrates the learned spatio-temporal dependencies to capture long-term complex spatio-temporal features.Considering that air pollution is related to the meteorological conditions of the city,the knowledge acquired from meteorological data is used in the model to enhance PM2.5 prediction performance.The input of the GAT-GRU model consists of PM2.5 concentration data and meteorological data.In order to verify the effectiveness of the proposed GAT-GRU prediction model,this paper designs experiments on real-world datasets compared with other baselines.Experimental results prove that our model achieves excellent performance in PM2.5 concentration prediction.
基金supported by the National Natural Science Foundation of China under Grant 62172059,61972057 and 62072055Hunan Provincial Natural Science Foundations of China under Grant 2020JJ4626+1 种基金Scientific Research Fund of Hunan Provincial Education Department of China under Grant 19B004Postgraduate Scientific Research Innovation Project of Hunan Province under Grant CX20210811.
文摘Image inpainting based on deep learning has been greatly improved.The original purpose of image inpainting was to repair some broken photos, suchas inpainting artifacts. However, it may also be used for malicious operations,such as destroying evidence. Therefore, detection and localization of imageinpainting operations are essential. Recent research shows that high-pass filteringfull convolutional network (HPFCN) is applied to image inpainting detection andachieves good results. However, those methods did not consider the spatial location and channel information of the feature map. To solve these shortcomings, weintroduce the squeezed excitation blocks (SE) and propose a high-pass filter attention full convolutional network (HPACN). In feature extraction, we apply concurrent spatial and channel attention (scSE) to enhance feature extraction and obtainmore information. Channel attention (cSE) is introduced in upsampling toenhance detection and localization. The experimental results show that the proposed method can achieve improvement on ImageNet.
基金This research has been supported by the US National Science Foundation under grant IIS-1115417the National Natural Science Foundation of China under grant 61728205,61472267and Foundation of Key Laboratory in Science and Technology Development Project of Suzhou under grant SZS201609。
文摘Multi-target regression is concerned with the simultaneous prediction of multiple continuous target variables based on the same set of input variables.It has received relatively small attention from the Machine Learning community.However,multi-target regression exists in many real-world applications.In this paper we conduct extensive experiments to investigate the performance of three representative multi-target regression learning algorithms(i.e.Multi-Target Stacking(MTS),Random Linear Target Combination(RLTC),and Multi-Objective Random Forest(MORF)),comparing the baseline single-target learning.Our experimental results show that all three multi-target regression learning algorithms do improve the performance of the single-target learning.Among them,MTS performs the best,followed by RLTC,followed by MORF.However,the single-target learning sometimes still performs very well,even the best.This analysis sheds the light on multi-target regression learning and indicates that the single-target learning is a competitive baseline for multi-target regression learning on multi-target domains.
基金This research has been supported by the National Natural Science Foundation of China under grant 61728205,61673290,61672371,61750110534,61876217Science&Technology Development Project of Suzhou under grant SYG201817.
文摘The pattern of thematic progression,reflecting the semantic relationships between contextual two sentences,is an important subject in discourse analysis.We introduce a new corpus of Chinese news discourses annotated with thematic progression information and explore some computational methods to automatically extracting the discourse structural features of simplified thematic progression pattern(STPP)between contextual sentences in a text.Furthermore,these features are used in a hybrid approach to a major discourse analysis task,Chinese coreference resolution.This novel approach is built up via heuristic sieves and a machine learning method that comprehensively utilizes both the top-down STPP features and the bottom-up semantic features.Experimental results on the intersection of the CoNLL-2012 task shared dataset and the CDTC corpus demonstrate the effectiveness of our proposed approach.
