As a large amount of data is increasingly generated from edge devices,such as smart homes,mobile phones,and wearable devices,it becomes crucial for many applications to deploy machine learning modes across edge device...As a large amount of data is increasingly generated from edge devices,such as smart homes,mobile phones,and wearable devices,it becomes crucial for many applications to deploy machine learning modes across edge devices.The execution speed of the deployed model is a key element to ensure service quality.Considering a highly heterogeneous edge deployment scenario,deep learning compiling is a novel approach that aims to solve this problem.It defines models using certain DSLs and generates efficient code implementations on different hardware devices.However,there are still two aspects that are not yet thoroughly investigated yet.The first is the optimization of memory-intensive operations,and the second problem is the heterogeneity of the deployment target.To that end,in this work,we propose a system solution that optimizes memory-intensive operation,optimizes the subgraph distribution,and enables the compiling and deployment of DNN models on multiple targets.The evaluation results show the performance of our proposed system.展开更多
[Objective]Accurate prediction of tomato growth height is crucial for optimizing production environments in smart farming.However,current prediction methods predominantly rely on empirical,mechanistic,or learning-base...[Objective]Accurate prediction of tomato growth height is crucial for optimizing production environments in smart farming.However,current prediction methods predominantly rely on empirical,mechanistic,or learning-based models that utilize either images data or environmental data.These methods fail to fully leverage multi-modal data to capture the diverse aspects of plant growth comprehensively.[Methods]To address this limitation,a two-stage phenotypic feature extraction(PFE)model based on deep learning algorithm of recurrent neural network(RNN)and long short-term memory(LSTM)was developed.The model integrated environment and plant information to provide a holistic understanding of the growth process,emploied phenotypic and temporal feature extractors to comprehensively capture both types of features,enabled a deeper understanding of the interaction between tomato plants and their environment,ultimately leading to highly accurate predictions of growth height.[Results and Discussions]The experimental results showed the model's ef‐fectiveness:When predicting the next two days based on the past five days,the PFE-based RNN and LSTM models achieved mean absolute percentage error(MAPE)of 0.81%and 0.40%,respectively,which were significantly lower than the 8.00%MAPE of the large language model(LLM)and 6.72%MAPE of the Transformer-based model.In longer-term predictions,the 10-day prediction for 4 days ahead and the 30-day prediction for 12 days ahead,the PFE-RNN model continued to outperform the other two baseline models,with MAPE of 2.66%and 14.05%,respectively.[Conclusions]The proposed method,which leverages phenotypic-temporal collaboration,shows great potential for intelligent,data-driven management of tomato cultivation,making it a promising approach for enhancing the efficiency and precision of smart tomato planting management.展开更多
A multi-bit antifuse-type one-time programmable （OTP） memory is designed, which has a smaller area and a shorter programming time compared with the conventional single-bit antifuse-type OTP memory. While the convent...A multi-bit antifuse-type one-time programmable （OTP） memory is designed, which has a smaller area and a shorter programming time compared with the conventional single-bit antifuse-type OTP memory. While the conventional antifuse-type OTP memory can store a bit per cell, a proposed OTP memory can store two consecutive bits per cell through a data compression technique. The 1 kbit OTP memory designed with Magnachip 0.18 μm CMOS （complementary metal-oxide semiconductor） process is 34% smaller than the conventional single-bit antifuse-type OTP memory since the sizes of cell array and row decoder are reduced. And the programming time of the proposed OTP memory is nearly 50% smaller than that of the conventional counterpart since two consecutive bytes can be compressed and programmed into eight OTP cells at once. The layout area is 214 μm× 327 μ,, and the read current is simulated to be 30.4 μA.展开更多
The information integration method of semantic web based on agent ontology(SWAO method) was put forward aiming at the problems in current network environment,which integrates,analyzes and processes enormous web inform...The information integration method of semantic web based on agent ontology(SWAO method) was put forward aiming at the problems in current network environment,which integrates,analyzes and processes enormous web information and extracts answers on the basis of semantics. With SWAO method as the clue,the following technologies were studied:the method of concept extraction based on semantic term mining,agent ontology construction method on account of multi-points and the answer extraction in view of semantic inference. Meanwhile,the structural model of the question answering system applying ontology was presented,which adopts OWL language to describe domain knowledge from where QA system infers and extracts answers by Jena inference engine. In the system testing,the precision rate reaches 86%,and the recalling rate is 93%. The experimental results prove that it is feasible to use the method to develop a question answering system,which is valuable for further study in more depth.展开更多
Reinforcement learning has been applied to air combat problems in recent years,and the idea of curriculum learning is often used for reinforcement learning,but traditional curriculum learning suffers from the problem ...Reinforcement learning has been applied to air combat problems in recent years,and the idea of curriculum learning is often used for reinforcement learning,but traditional curriculum learning suffers from the problem of plasticity loss in neural networks.Plasticity loss is the difficulty of learning new knowledge after the network has converged.To this end,we propose a motivational curriculum learning distributed proximal policy optimization(MCLDPPO)algorithm,through which trained agents can significantly outperform the predictive game tree and mainstream reinforcement learning methods.The motivational curriculum learning is designed to help the agent gradually improve its combat ability by observing the agent's unsatisfactory performance and providing appropriate rewards as a guide.Furthermore,a complete tactical maneuver is encapsulated based on the existing air combat knowledge,and through the flexible use of these maneuvers,some tactics beyond human knowledge can be realized.In addition,we designed an interruption mechanism for the agent to increase the frequency of decisionmaking when the agent faces an emergency.When the number of threats received by the agent changes,the current action is interrupted in order to reacquire observations and make decisions again.Using the interruption mechanism can significantly improve the performance of the agent.To simulate actual air combat better,we use digital twin technology to simulate real air battles and propose a parallel battlefield mechanism that can run multiple simulation environments simultaneously,effectively improving data throughput.The experimental results demonstrate that the agent can fully utilize the situational information to make reasonable decisions and provide tactical adaptation in the air combat,verifying the effectiveness of the algorithmic framework proposed in this paper.展开更多
Evolutionary algorithms(EAs)have been used in high utility itemset mining(HUIM)to address the problem of discover-ing high utility itemsets(HUIs)in the exponential search space.EAs have good running and mining perform...Evolutionary algorithms(EAs)have been used in high utility itemset mining(HUIM)to address the problem of discover-ing high utility itemsets(HUIs)in the exponential search space.EAs have good running and mining performance,but they still require huge computational resource and may miss many HUIs.Due to the good combination of EA and graphics processing unit(GPU),we propose a parallel genetic algorithm(GA)based on the platform of GPU for mining HUIM(PHUI-GA).The evolution steps with improvements are performed in central processing unit(CPU)and the CPU intensive steps are sent to GPU to eva-luate with multi-threaded processors.Experiments show that the mining performance of PHUI-GA outperforms the existing EAs.When mining 90%HUIs,the PHUI-GA is up to 188 times better than the existing EAs and up to 36 times better than the CPU parallel approach.展开更多
Interacting multiple models is the hotspot in the research of maneuvering target models at present. A hierarchical idea is introduced into IMM algorithm. The method is that the whole models are organized as two levels...Interacting multiple models is the hotspot in the research of maneuvering target models at present. A hierarchical idea is introduced into IMM algorithm. The method is that the whole models are organized as two levels to co-work, and each cell model is an improved ＂current＂ statistical model. In the improved model, a kind of nonlinear fuzzy membership function is presented to get over the limitation of original model, which can not track weak maneuvering target precisely. At last, simulation experiments prove the efficient of the novel algorithm compared to interacting multiple model and hierarchical interacting multiple model based original ＂current＂ statistical model in tracking precision.展开更多
In order to enhance the robustness and contrast in the minimum variance(MV) beamformer, adaptive diagonal loading method was proposed. The conventional diagonal loading technique has already been used in the MV beamfo...In order to enhance the robustness and contrast in the minimum variance(MV) beamformer, adaptive diagonal loading method was proposed. The conventional diagonal loading technique has already been used in the MV beamformer, but has the drawback that its level is specified by predefined parameter and without consideration of input-data. To alleviate this problem, the level of diagonal loading was computed appropriately and automatically from the given data by shrinkage method in the proposed adaptive diagonal loaded beamformer. The performance of the proposed beamformer was tested on the simulated point target and cyst phantom was obtained using Field II. In the point target simulation, it is shown that the proposed method has higher lateral resolution than the conventional delay-and-sum beamformer and could be more robust in estimating the amplitude peak than the MV beamformer when acoustic velocity error exists. In the cyst phantom simulation, the proposed beamformer has shown that it achieves an improvement in contrast ratio and without distorting the edges of cyst.展开更多
The setting of attention parameters plays a role in the performance of synergetic neural network based on PFAP model. This paper first analyzes the attention parameter setting algorithm based on award-penalty learning...The setting of attention parameters plays a role in the performance of synergetic neural network based on PFAP model. This paper first analyzes the attention parameter setting algorithm based on award-penalty learning mechanism. Then, it presents an improved algorithm to overcome its drawbacks. The experimental results demonstrate that the novel algorithm is better than the original one under the same circumstances.展开更多
Based on the rough set theory which is a powerful tool in dealing with vagueness and uncertainty, an algorithm to mine association rules in incomplete information systems was presented and the support and confidence w...Based on the rough set theory which is a powerful tool in dealing with vagueness and uncertainty, an algorithm to mine association rules in incomplete information systems was presented and the support and confidence were redefined. The algorithm can mine the association rules with decision attributes directly without processing missing values. Using the incomplete dataset Mushroom from UCI machine learning repository, the new algorithm was compared with the classical association rules mining algorithm based on Apriori from the number of rules extracted, testing accuracy and execution time. The experiment results show that the new algorithm has advantages of short execution time and high accuracy.展开更多
The basic"current"statistical model and adaptive Kalman filter algorithm can not track a weakly maneuvering target precisely,though it has good estimate accuracy for strongly maneuvering target.In order to s...The basic"current"statistical model and adaptive Kalman filter algorithm can not track a weakly maneuvering target precisely,though it has good estimate accuracy for strongly maneuvering target.In order to solve this problem,a novel nonlinear fuzzy membership function was presented to adjust the upper and lower limit of target acceleration adaptively,and then the validity of the new algorithm for feeblish maneuvering target was proved in theory.At last,the computer simulation experiments indicated that the new algorithm has a great advantage over the basic"current"statistical model and adaptive algorithm.展开更多
With the development of the support vector machine（SVM）,the kernel function has become one of the cores of the research on SVM.To a large extent,the kernel function determines the generalization ability of the class...With the development of the support vector machine（SVM）,the kernel function has become one of the cores of the research on SVM.To a large extent,the kernel function determines the generalization ability of the classifier,but there is still no general theory to guide the choice and structure of the kernel function.An ensemble kernel function model based on the game theory is proposed,which is used for the SVM classification algorithm.The model can effectively integrate the advantages of the local kernel and the global kernel to get a better classification result,and can provide a feasible way for structuring the kernel function.By making experiments on some standard datasets,it is verified that the new method can significantly improve the accuracy of classification.展开更多
In sensor networks,the adversaries can inject false data reports from compromised nodes.Previous approaches to filter false reports,e.g.,SEF,only verify the correctness of the message authentication code (MACs) carrie...In sensor networks,the adversaries can inject false data reports from compromised nodes.Previous approaches to filter false reports,e.g.,SEF,only verify the correctness of the message authentication code (MACs) carried in each data report on intermediate nodes,thus cannot filter out fake reports that are forged in a collaborative manner by a group of compromised nodes,even if these compromised nodes distribute in different geographical areas.Furthermore,if the adversary obtains keys from enough (e.g.,more than t in SEF) distinct key partitions,it then can successfully forge a data report without being detected en-route.A neighbor information based false report filtering scheme (NFFS) in wireless sensor networks was presented.In NFFS,each node distributes its neighbor information to some other nodes after deployment.When a report is generated for an observed event,it must carry the IDs and the MACs from t detecting nodes.Each forwarding node checks not only the correctness of the MACs carried in the report,but also the legitimacy of the relative position of these detecting nodes.Analysis and simulation results demonstrate that NFFS can resist collaborative false data injection attacks efficiently,and thus can tolerate much more compromised nodes than existing schemes.展开更多
基金supported by the National Natural Science Foundation of China(U21A20519)。
文摘As a large amount of data is increasingly generated from edge devices,such as smart homes,mobile phones,and wearable devices,it becomes crucial for many applications to deploy machine learning modes across edge devices.The execution speed of the deployed model is a key element to ensure service quality.Considering a highly heterogeneous edge deployment scenario,deep learning compiling is a novel approach that aims to solve this problem.It defines models using certain DSLs and generates efficient code implementations on different hardware devices.However,there are still two aspects that are not yet thoroughly investigated yet.The first is the optimization of memory-intensive operations,and the second problem is the heterogeneity of the deployment target.To that end,in this work,we propose a system solution that optimizes memory-intensive operation,optimizes the subgraph distribution,and enables the compiling and deployment of DNN models on multiple targets.The evaluation results show the performance of our proposed system.
文摘[Objective]Accurate prediction of tomato growth height is crucial for optimizing production environments in smart farming.However,current prediction methods predominantly rely on empirical,mechanistic,or learning-based models that utilize either images data or environmental data.These methods fail to fully leverage multi-modal data to capture the diverse aspects of plant growth comprehensively.[Methods]To address this limitation,a two-stage phenotypic feature extraction(PFE)model based on deep learning algorithm of recurrent neural network(RNN)and long short-term memory(LSTM)was developed.The model integrated environment and plant information to provide a holistic understanding of the growth process,emploied phenotypic and temporal feature extractors to comprehensively capture both types of features,enabled a deeper understanding of the interaction between tomato plants and their environment,ultimately leading to highly accurate predictions of growth height.[Results and Discussions]The experimental results showed the model's ef‐fectiveness:When predicting the next two days based on the past five days,the PFE-based RNN and LSTM models achieved mean absolute percentage error(MAPE)of 0.81%and 0.40%,respectively,which were significantly lower than the 8.00%MAPE of the large language model(LLM)and 6.72%MAPE of the Transformer-based model.In longer-term predictions,the 10-day prediction for 4 days ahead and the 30-day prediction for 12 days ahead,the PFE-RNN model continued to outperform the other two baseline models,with MAPE of 2.66%and 14.05%,respectively.[Conclusions]The proposed method,which leverages phenotypic-temporal collaboration,shows great potential for intelligent,data-driven management of tomato cultivation,making it a promising approach for enhancing the efficiency and precision of smart tomato planting management.
基金Project supported by the 2nd Stage of Brain KoreaProject supported by the Korea Research Foundation
文摘A multi-bit antifuse-type one-time programmable （OTP） memory is designed, which has a smaller area and a shorter programming time compared with the conventional single-bit antifuse-type OTP memory. While the conventional antifuse-type OTP memory can store a bit per cell, a proposed OTP memory can store two consecutive bits per cell through a data compression technique. The 1 kbit OTP memory designed with Magnachip 0.18 μm CMOS （complementary metal-oxide semiconductor） process is 34% smaller than the conventional single-bit antifuse-type OTP memory since the sizes of cell array and row decoder are reduced. And the programming time of the proposed OTP memory is nearly 50% smaller than that of the conventional counterpart since two consecutive bytes can be compressed and programmed into eight OTP cells at once. The layout area is 214 μm× 327 μ,, and the read current is simulated to be 30.4 μA.
基金Supported by National Basic Research Program （973 Program） of China （2007CB724205）, National Natural Science Foundation of China （60604010）, and China Postdoctoral Science Foundation Funded Project （20080440384）
基金Projects(60773462, 60672171) supported by the National Natural Science Foundation of ChinaProjects(2009AA12143, 2009AA012136) supported by the National High-Tech Research and Development Program of ChinaProject(20080430250) supported by the Foundation of Post-Doctor in China
文摘The information integration method of semantic web based on agent ontology(SWAO method) was put forward aiming at the problems in current network environment,which integrates,analyzes and processes enormous web information and extracts answers on the basis of semantics. With SWAO method as the clue,the following technologies were studied:the method of concept extraction based on semantic term mining,agent ontology construction method on account of multi-points and the answer extraction in view of semantic inference. Meanwhile,the structural model of the question answering system applying ontology was presented,which adopts OWL language to describe domain knowledge from where QA system infers and extracts answers by Jena inference engine. In the system testing,the precision rate reaches 86%,and the recalling rate is 93%. The experimental results prove that it is feasible to use the method to develop a question answering system,which is valuable for further study in more depth.
基金Supported by National Natural Science Foundation of China （60503024, 60375038, 60374032） and 0pen Fund of Key Laboratory of Industrial Controlling Technology, Zhejiang Universlty （060004）
文摘Reinforcement learning has been applied to air combat problems in recent years,and the idea of curriculum learning is often used for reinforcement learning,but traditional curriculum learning suffers from the problem of plasticity loss in neural networks.Plasticity loss is the difficulty of learning new knowledge after the network has converged.To this end,we propose a motivational curriculum learning distributed proximal policy optimization(MCLDPPO)algorithm,through which trained agents can significantly outperform the predictive game tree and mainstream reinforcement learning methods.The motivational curriculum learning is designed to help the agent gradually improve its combat ability by observing the agent's unsatisfactory performance and providing appropriate rewards as a guide.Furthermore,a complete tactical maneuver is encapsulated based on the existing air combat knowledge,and through the flexible use of these maneuvers,some tactics beyond human knowledge can be realized.In addition,we designed an interruption mechanism for the agent to increase the frequency of decisionmaking when the agent faces an emergency.When the number of threats received by the agent changes,the current action is interrupted in order to reacquire observations and make decisions again.Using the interruption mechanism can significantly improve the performance of the agent.To simulate actual air combat better,we use digital twin technology to simulate real air battles and propose a parallel battlefield mechanism that can run multiple simulation environments simultaneously,effectively improving data throughput.The experimental results demonstrate that the agent can fully utilize the situational information to make reasonable decisions and provide tactical adaptation in the air combat,verifying the effectiveness of the algorithmic framework proposed in this paper.
基金This work was supported by the National Natural Science Foundation of China(62073155,62002137,62106088,62206113)the High-End Foreign Expert Recruitment Plan(G2023144007L)the Fundamental Research Funds for the Central Universities(JUSRP221028).
文摘Evolutionary algorithms(EAs)have been used in high utility itemset mining(HUIM)to address the problem of discover-ing high utility itemsets(HUIs)in the exponential search space.EAs have good running and mining performance,but they still require huge computational resource and may miss many HUIs.Due to the good combination of EA and graphics processing unit(GPU),we propose a parallel genetic algorithm(GA)based on the platform of GPU for mining HUIM(PHUI-GA).The evolution steps with improvements are performed in central processing unit(CPU)and the CPU intensive steps are sent to GPU to eva-luate with multi-threaded processors.Experiments show that the mining performance of PHUI-GA outperforms the existing EAs.When mining 90%HUIs,the PHUI-GA is up to 188 times better than the existing EAs and up to 36 times better than the CPU parallel approach.
文摘Interacting multiple models is the hotspot in the research of maneuvering target models at present. A hierarchical idea is introduced into IMM algorithm. The method is that the whole models are organized as two levels to co-work, and each cell model is an improved ＂current＂ statistical model. In the improved model, a kind of nonlinear fuzzy membership function is presented to get over the limitation of original model, which can not track weak maneuvering target precisely. At last, simulation experiments prove the efficient of the novel algorithm compared to interacting multiple model and hierarchical interacting multiple model based original ＂current＂ statistical model in tracking precision.
基金Project(2013GZX0147-3)supported by the Science and Technology Pillar Program of Sichuan Province,China
文摘In order to enhance the robustness and contrast in the minimum variance(MV) beamformer, adaptive diagonal loading method was proposed. The conventional diagonal loading technique has already been used in the MV beamformer, but has the drawback that its level is specified by predefined parameter and without consideration of input-data. To alleviate this problem, the level of diagonal loading was computed appropriately and automatically from the given data by shrinkage method in the proposed adaptive diagonal loaded beamformer. The performance of the proposed beamformer was tested on the simulated point target and cyst phantom was obtained using Field II. In the point target simulation, it is shown that the proposed method has higher lateral resolution than the conventional delay-and-sum beamformer and could be more robust in estimating the amplitude peak than the MV beamformer when acoustic velocity error exists. In the cyst phantom simulation, the proposed beamformer has shown that it achieves an improvement in contrast ratio and without distorting the edges of cyst.
基金Supported by Fujian Natural Science Foundation(A0110010).
文摘The setting of attention parameters plays a role in the performance of synergetic neural network based on PFAP model. This paper first analyzes the attention parameter setting algorithm based on award-penalty learning mechanism. Then, it presents an improved algorithm to overcome its drawbacks. The experimental results demonstrate that the novel algorithm is better than the original one under the same circumstances.
基金Projects(10871031, 60474070) supported by the National Natural Science Foundation of ChinaProject(07A001) supported by the Scientific Research Fund of Hunan Provincial Education Department, China
文摘Based on the rough set theory which is a powerful tool in dealing with vagueness and uncertainty, an algorithm to mine association rules in incomplete information systems was presented and the support and confidence were redefined. The algorithm can mine the association rules with decision attributes directly without processing missing values. Using the incomplete dataset Mushroom from UCI machine learning repository, the new algorithm was compared with the classical association rules mining algorithm based on Apriori from the number of rules extracted, testing accuracy and execution time. The experiment results show that the new algorithm has advantages of short execution time and high accuracy.
文摘The basic"current"statistical model and adaptive Kalman filter algorithm can not track a weakly maneuvering target precisely,though it has good estimate accuracy for strongly maneuvering target.In order to solve this problem,a novel nonlinear fuzzy membership function was presented to adjust the upper and lower limit of target acceleration adaptively,and then the validity of the new algorithm for feeblish maneuvering target was proved in theory.At last,the computer simulation experiments indicated that the new algorithm has a great advantage over the basic"current"statistical model and adaptive algorithm.
基金supported by the National Natural Science Foundation of China(U1433116)the Aviation Science Foundation of China(20145752033)the Graduate Innovation Project of Jiangsu Province(KYLX15_0324)
文摘With the development of the support vector machine（SVM）,the kernel function has become one of the cores of the research on SVM.To a large extent,the kernel function determines the generalization ability of the classifier,but there is still no general theory to guide the choice and structure of the kernel function.An ensemble kernel function model based on the game theory is proposed,which is used for the SVM classification algorithm.The model can effectively integrate the advantages of the local kernel and the global kernel to get a better classification result,and can provide a feasible way for structuring the kernel function.By making experiments on some standard datasets,it is verified that the new method can significantly improve the accuracy of classification.
基金Projects(61173169,61103203,70921001)supported by the National Natural Science Foundation of ChinaProject(NCET-10-0798)supported by Program for New Century Excellent Talents in University of China
文摘In sensor networks,the adversaries can inject false data reports from compromised nodes.Previous approaches to filter false reports,e.g.,SEF,only verify the correctness of the message authentication code (MACs) carried in each data report on intermediate nodes,thus cannot filter out fake reports that are forged in a collaborative manner by a group of compromised nodes,even if these compromised nodes distribute in different geographical areas.Furthermore,if the adversary obtains keys from enough (e.g.,more than t in SEF) distinct key partitions,it then can successfully forge a data report without being detected en-route.A neighbor information based false report filtering scheme (NFFS) in wireless sensor networks was presented.In NFFS,each node distributes its neighbor information to some other nodes after deployment.When a report is generated for an observed event,it must carry the IDs and the MACs from t detecting nodes.Each forwarding node checks not only the correctness of the MACs carried in the report,but also the legitimacy of the relative position of these detecting nodes.Analysis and simulation results demonstrate that NFFS can resist collaborative false data injection attacks efficiently,and thus can tolerate much more compromised nodes than existing schemes.
