Recurrent neural networks (RNNs) have been heavily used in applications relying on sequence data such as time series and natural languages. As a matter of fact, their behaviors lack rigorous quality assurance due to t...Recurrent neural networks (RNNs) have been heavily used in applications relying on sequence data such as time series and natural languages. As a matter of fact, their behaviors lack rigorous quality assurance due to the black-box nature of deep learning. It is an urgent and challenging task to formally reason about the behaviors of RNNs. To this end, we first present an extension of linear-time temporal logic to reason about properties with respect to RNNs, such as local robustness, reachability, and some temporal properties. Based on the proposed logic, we formalize the verification obligation as a Hoare-like triple, from both qualitative and quantitative perspectives. The former concerns whether all the outputs resulting from the inputs fulfilling the pre-condition satisfy the post-condition, whereas the latter is to compute the probability that the post-condition is satisfied on the premise that the inputs fulfill the pre-condition. To tackle these problems, we develop a systematic verification framework, mainly based on polyhedron propagation, dimension-preserving abstraction, and the Monte Carlo sampling. We also implement our algorithm with a prototype tool and conduct experiments to demonstrate its feasibility and efficiency.展开更多
Laser-induced breakdown spectroscopy(LIBS)has been applied to many fields for the quantitative analysis of diverse materials.Improving the prediction accuracy of LIBS regression models is still of great significance f...Laser-induced breakdown spectroscopy(LIBS)has been applied to many fields for the quantitative analysis of diverse materials.Improving the prediction accuracy of LIBS regression models is still of great significance for the Mars exploration in the near future.In this study,we explored the quantitative analysis of LIBS for the one-dimensional Chem Cam(an instrument containing a LIBS spectrometer and a Remote Micro-Imager)spectral data whose spectra are produced by the Chem Cam team using LIBS under the Mars-like atmospheric conditions.We constructed a convolutional neural network(CNN)regression model with unified parameters for all oxides,which is efficient and concise.CNN that has the excellent capability of feature extraction can effectively overcome the chemical matrix effects that impede the prediction accuracy of regression models.Firstly,we explored the effects of four activation functions on the performance of the CNN model.The results show that the CNN model with the hyperbolic tangent(tanh)function outperforms the CNN models with the other activation functions(the rectified linear unit function,the linear function and the Sigmoid function).Secondly,we compared the performance among the CNN models using different optimization methods.The CNN model with the stochastic gradient descent optimization and the initial learning rate?=?0.0005 achieves satisfactory performance compared to the other CNN models.Finally,we compared the performance of the CNN model,the model based on support vector regression(SVR)and the model based on partial least square regression(PLSR).The results exhibit the CNN model is superior to the SVR model and the PLSR model for all oxides.Based on the above analysis,we conclude the CNN regression model can effectively improve the prediction accuracy of LIBS.展开更多
The distribution characteristics of rare earth elements （REE） in bottomsediments are influenced by many factors. Hence, conducting a quantitative analysis isdifficult. A qualitative analysis of the relationships bet...The distribution characteristics of rare earth elements （REE） in bottomsediments are influenced by many factors. Hence, conducting a quantitative analysis isdifficult. A qualitative analysis of the relationships between ΣREE content andprovenance, hydrodynamics, grain size and mineral distribution in the Beibu Gulf showsthat terrestrial rocks control the ΣREE composition. Both weaker hydrodynamics andfiner grain size lead to a higher ΣREE content. Relative curves revealing therelationships between individual impact factors and ΣREE content were obtained fromthe combination of qualitative and quantitative analyses of the BP neural network,which trained the position of samples, gravel content, sand content, silt content, claycontent and clay mineral content. The results are consistent with those of thequantitative analysis. The self-learning algorithm is automatically determined andcalculated quantitatively. The impact of each factor on REEs and how each factorcontrols the ΣREE distribution is identified. Thus, environmental changes and thegeological evolution of the region can be inferred based on curve variation and the geological evolution of the region can be inferred based on curve variation and theactual situation. This method also provides useful theoretical guidance for the analysisof REE enrichment and dispersion.展开更多
In energy dispersive X-ray fiuorescence(EDXRF), quantitative elemental content analysis becomes difficult due to the existence of the noise, the spectrum peak superposition, element matrix effect, etc. In this paper, ...In energy dispersive X-ray fiuorescence(EDXRF), quantitative elemental content analysis becomes difficult due to the existence of the noise, the spectrum peak superposition, element matrix effect, etc. In this paper, a hybrid approach of genetic algorithm(GA) and back propagation(BP) neural network is proposed without considering the complex relationship between the elemental content and peak intensity. The aim of GA-optimized BP is to get better network initial weights and thresholds. The starting point of this approach is that the reciprocal of the mean square error of the initialization BP neural network is set as the fitness value of the individuals in GA; and the initial weights and thresholds are replaced by individuals, then the optimal individual is searched by selecting, crossover and mutation operations, finally a new BP neural network model is established with the optimal initial weights and thresholds. The quantitative analysis results of titanium and iron contents in five types of mineral samples show that the relative errors of 76.7% samples are below 2%, compared to chemical analysis data, which demonstrates the effectiveness of the proposed method.展开更多
BP and RBF neural network to predict forest stock volume were studied,but the study in evaluating both networks’ application effects was not conducted.In order to find a higher forecast precision,more strong applicat...BP and RBF neural network to predict forest stock volume were studied,but the study in evaluating both networks’ application effects was not conducted.In order to find a higher forecast precision,more strong applicative method,the comprehensive analysis and evaluation on the two methods were carried out in the practical application. By the correlation analysis,crown density,shady-slope and sunny-slope,TM1,TM2,TM3,TM5, TM7,NDVI,TM,(4-3),TM4/3 were selected as input variables,and the forest volume of Miyun County as output variables,RBF and BP neural network models for forecasting the forest volume were established.And the neural network training step length,training time,prediction accuracy and the applicability model of the two methods were comprehensively analyzed.The results show that the RBF neural network model is superior to the BP neural network model.展开更多
Medical image analysis has become a cornerstone of modern healthcare,driven by the exponential growth of data from imaging modalities such as MRI,CT,PET,ultrasound,and X-ray.Traditional machine learning methods have m...Medical image analysis has become a cornerstone of modern healthcare,driven by the exponential growth of data from imaging modalities such as MRI,CT,PET,ultrasound,and X-ray.Traditional machine learning methods have made early contributions;however,recent advancements in deep learning(DL)have revolutionized the field,offering state-of-the-art performance in image classification,segmentation,detection,fusion,registration,and enhancement.This comprehensive review presents an in-depth analysis of deep learning methodologies applied across medical image analysis tasks,highlighting both foundational models and recent innovations.The article begins by introducing conventional techniques and their limitations,setting the stage for DL-based solutions.Core DL architectures,including Convolutional Neural Networks(CNNs),Recurrent Neural Networks(RNNs),Generative Adversarial Networks(GANs),Vision Transformers(ViTs),and hybrid models,are discussed in detail,including their advantages and domain-specific adaptations.Advanced learning paradigms such as semi-supervised learning,selfsupervised learning,and few-shot learning are explored for their potential to mitigate data annotation challenges in clinical datasets.This review further categorizes major tasks in medical image analysis,elaborating on how DL techniques have enabled precise tumor segmentation,lesion detection,modality fusion,super-resolution,and robust classification across diverse clinical settings.Emphasis is placed on applications in oncology,cardiology,neurology,and infectious diseases,including COVID-19.Challenges such as data scarcity,label imbalance,model generalizability,interpretability,and integration into clinical workflows are critically examined.Ethical considerations,explainable AI(XAI),federated learning,and regulatory compliance are discussed as essential components of real-world deployment.Benchmark datasets,evaluation metrics,and comparative performance analyses are presented to support future research.The article concludes with a forward-looking perspective on the role of foundation models,multimodal learning,edge AI,and bio-inspired computing in the future of medical imaging.Overall,this review serves as a valuable resource for researchers,clinicians,and developers aiming to harness deep learning for intelligent,efficient,and clinically viable medical image analysis.展开更多
The condition monitoring and fault diagnosis of rolling element bearings are particularly crucial in rotating mechanical applications in industry. A bearing fault signal contains information not only about fault condi...The condition monitoring and fault diagnosis of rolling element bearings are particularly crucial in rotating mechanical applications in industry. A bearing fault signal contains information not only about fault condition and fault type but also the severity of the fault. This means fault severity quantitative analysis is one of most active and valid ways to realize proper maintenance decision. Aiming at the deficiency of the research in bearing single point pitting fault quantitative diagnosis, a new back-propagation neural network method based on wavelet packet decomposition coefficient entropy is proposed. The three levels of wavelet packet coefficient entropy（WPCE） is introduced as a characteristic input vector to the BPNN. Compared with the wavelet packet decomposition energy ratio input vector, WPCE shows more sensitive in distinguishing from the different fault severity degree of the measured signal. The engineering application results show that the quantitative trend fault diagnosis is realized in the different fault degree of the single point bearing pitting fault. The breakthrough attempt from quantitative to qualitative on the pattern recognition of rolling element bearings fault diagnosis is realized.展开更多
The dynamic,heterogeneous nature of Edge computing in the Internet of Things(Edge-IoT)and Industrial IoT(IIoT)networks brings unique and evolving cybersecurity challenges.This study maps cyber threats in Edge-IoT/IIoT...The dynamic,heterogeneous nature of Edge computing in the Internet of Things(Edge-IoT)and Industrial IoT(IIoT)networks brings unique and evolving cybersecurity challenges.This study maps cyber threats in Edge-IoT/IIoT environments to the Adversarial Tactics,Techniques,and Common Knowledge(ATT&CK)framework by MITRE and introduces a lightweight,data-driven scoring model that enables rapid identification and prioritization of attacks.Inspired by the Factor Analysis of Information Risk model,our proposed scoring model integrates four key metrics:Common Vulnerability Scoring System(CVSS)-based severity scoring,Cyber Kill Chain–based difficulty estimation,Deep Neural Networks-driven detection scoring,and frequency analysis based on dataset prevalence.By aggregating these indicators,the model generates comprehensive risk profiles,facilitating actionable prioritization of threats.Robustness and stability of the scoring model are validated through non-parametric correlation analysis using Spearman’s and Kendall’s rank correlation coefficients,demonstrating consistent performance across diverse scenarios.The approach culminates in a prioritized attack ranking that provides actionable guidance for risk mitigation and resource allocation in Edge-IoT/IIoT security operations.By leveraging real-world data to align MITRE ATT&CK techniques with CVSS metrics,the framework offers a standardized and practically applicable solution for consistent threat assessment in operational settings.The proposed lightweight scoring model delivers rapid and reliable results under dynamic cyber conditions,facilitating timely identification of attack scenarios and prioritization of response strategies.Our systematic integration of established taxonomies with data-driven indicators strengthens practical risk management and supports strategic planning in next-generation IoT deployments.Ultimately,this work advances adaptive threat modeling for Edge/IIoT ecosystems and establishes a robust foundation for evidence-based prioritization in emerging cyber-physical infrastructures.展开更多
Qualitative spacecraft pursuit-evasion problem which focuses on feasibility is rarely studied because of high-dimensional dynamics,intractable terminal constraints and heavy computational cost.In this paper,A physics-...Qualitative spacecraft pursuit-evasion problem which focuses on feasibility is rarely studied because of high-dimensional dynamics,intractable terminal constraints and heavy computational cost.In this paper,A physics-informed framework is proposed for the problem,providing an intuitive method for spacecraft threat relationship determination,situation assessment,mission feasibility analysis and orbital game rules summarization.For the first time,situation adjustment suggestions can be provided for the weak player in orbital game.First,a dimension-reduction dynamics is derived in the line-of-sight rotation coordinate system and the qualitative model is determined,reducing complexity and avoiding the difficulty of target set presentation caused by individual modeling.Second,the Backwards Reachable Set(BRS)of the target set is used for state space partition and capture zone presentation.Reverse-time analysis can eliminate the influence of changeable initial state and enable the proposed framework to analyze plural situations simultaneously.Third,a time-dependent Hamilton-Jacobi-Isaacs(HJI)Partial Differential Equation(PDE)is established to describe BRS evolution driven by dimension-reduction dynamics,based on level set method.Then,Physics-Informed Neural Networks(PINNs)are extended to HJI PDE final value problem,supporting orbital game rules summarization through capture zone evolution analysis.Finally,numerical results demonstrate the feasibility and efficiency of the proposed framework.展开更多
基金supported by the National Natural Science Foundation of China under Grant Nos.61872371,62032024,and U19A2062the Open Fund from the State Key Laboratory of High Performance Computing of China(HPCL)under Grant No.202001-07.
文摘Recurrent neural networks (RNNs) have been heavily used in applications relying on sequence data such as time series and natural languages. As a matter of fact, their behaviors lack rigorous quality assurance due to the black-box nature of deep learning. It is an urgent and challenging task to formally reason about the behaviors of RNNs. To this end, we first present an extension of linear-time temporal logic to reason about properties with respect to RNNs, such as local robustness, reachability, and some temporal properties. Based on the proposed logic, we formalize the verification obligation as a Hoare-like triple, from both qualitative and quantitative perspectives. The former concerns whether all the outputs resulting from the inputs fulfilling the pre-condition satisfy the post-condition, whereas the latter is to compute the probability that the post-condition is satisfied on the premise that the inputs fulfill the pre-condition. To tackle these problems, we develop a systematic verification framework, mainly based on polyhedron propagation, dimension-preserving abstraction, and the Monte Carlo sampling. We also implement our algorithm with a prototype tool and conduct experiments to demonstrate its feasibility and efficiency.
基金supported by the Pre-research project on Civil Aerospace Technologies(No.D020102)funded by China National Space Administration(CNSA)the funding from National Natural Science Foundation of China(Nos.U1931211,41573056)+1 种基金the Natural Science Foundation of Shandong Province(No.ZR2019MD008)the Major Research Project of Shandong Province(No.GG201809130208)。
文摘Laser-induced breakdown spectroscopy(LIBS)has been applied to many fields for the quantitative analysis of diverse materials.Improving the prediction accuracy of LIBS regression models is still of great significance for the Mars exploration in the near future.In this study,we explored the quantitative analysis of LIBS for the one-dimensional Chem Cam(an instrument containing a LIBS spectrometer and a Remote Micro-Imager)spectral data whose spectra are produced by the Chem Cam team using LIBS under the Mars-like atmospheric conditions.We constructed a convolutional neural network(CNN)regression model with unified parameters for all oxides,which is efficient and concise.CNN that has the excellent capability of feature extraction can effectively overcome the chemical matrix effects that impede the prediction accuracy of regression models.Firstly,we explored the effects of four activation functions on the performance of the CNN model.The results show that the CNN model with the hyperbolic tangent(tanh)function outperforms the CNN models with the other activation functions(the rectified linear unit function,the linear function and the Sigmoid function).Secondly,we compared the performance among the CNN models using different optimization methods.The CNN model with the stochastic gradient descent optimization and the initial learning rate?=?0.0005 achieves satisfactory performance compared to the other CNN models.Finally,we compared the performance of the CNN model,the model based on support vector regression(SVR)and the model based on partial least square regression(PLSR).The results exhibit the CNN model is superior to the SVR model and the PLSR model for all oxides.Based on the above analysis,we conclude the CNN regression model can effectively improve the prediction accuracy of LIBS.
文摘The distribution characteristics of rare earth elements （REE） in bottomsediments are influenced by many factors. Hence, conducting a quantitative analysis isdifficult. A qualitative analysis of the relationships between ΣREE content andprovenance, hydrodynamics, grain size and mineral distribution in the Beibu Gulf showsthat terrestrial rocks control the ΣREE composition. Both weaker hydrodynamics andfiner grain size lead to a higher ΣREE content. Relative curves revealing therelationships between individual impact factors and ΣREE content were obtained fromthe combination of qualitative and quantitative analyses of the BP neural network,which trained the position of samples, gravel content, sand content, silt content, claycontent and clay mineral content. The results are consistent with those of thequantitative analysis. The self-learning algorithm is automatically determined andcalculated quantitatively. The impact of each factor on REEs and how each factorcontrols the ΣREE distribution is identified. Thus, environmental changes and thegeological evolution of the region can be inferred based on curve variation and the geological evolution of the region can be inferred based on curve variation and theactual situation. This method also provides useful theoretical guidance for the analysisof REE enrichment and dispersion.
基金Supported by National Outstanding Youth Science Foundation of China(No.41025015)the National Natural Science Foundation of China(No.41274109)Sichuan Youth Science and Technology Innovation Research Team(No.2011JTD0013)
文摘In energy dispersive X-ray fiuorescence(EDXRF), quantitative elemental content analysis becomes difficult due to the existence of the noise, the spectrum peak superposition, element matrix effect, etc. In this paper, a hybrid approach of genetic algorithm(GA) and back propagation(BP) neural network is proposed without considering the complex relationship between the elemental content and peak intensity. The aim of GA-optimized BP is to get better network initial weights and thresholds. The starting point of this approach is that the reciprocal of the mean square error of the initialization BP neural network is set as the fitness value of the individuals in GA; and the initial weights and thresholds are replaced by individuals, then the optimal individual is searched by selecting, crossover and mutation operations, finally a new BP neural network model is established with the optimal initial weights and thresholds. The quantitative analysis results of titanium and iron contents in five types of mineral samples show that the relative errors of 76.7% samples are below 2%, compared to chemical analysis data, which demonstrates the effectiveness of the proposed method.
文摘BP and RBF neural network to predict forest stock volume were studied,but the study in evaluating both networks’ application effects was not conducted.In order to find a higher forecast precision,more strong applicative method,the comprehensive analysis and evaluation on the two methods were carried out in the practical application. By the correlation analysis,crown density,shady-slope and sunny-slope,TM1,TM2,TM3,TM5, TM7,NDVI,TM,(4-3),TM4/3 were selected as input variables,and the forest volume of Miyun County as output variables,RBF and BP neural network models for forecasting the forest volume were established.And the neural network training step length,training time,prediction accuracy and the applicability model of the two methods were comprehensively analyzed.The results show that the RBF neural network model is superior to the BP neural network model.
文摘Medical image analysis has become a cornerstone of modern healthcare,driven by the exponential growth of data from imaging modalities such as MRI,CT,PET,ultrasound,and X-ray.Traditional machine learning methods have made early contributions;however,recent advancements in deep learning(DL)have revolutionized the field,offering state-of-the-art performance in image classification,segmentation,detection,fusion,registration,and enhancement.This comprehensive review presents an in-depth analysis of deep learning methodologies applied across medical image analysis tasks,highlighting both foundational models and recent innovations.The article begins by introducing conventional techniques and their limitations,setting the stage for DL-based solutions.Core DL architectures,including Convolutional Neural Networks(CNNs),Recurrent Neural Networks(RNNs),Generative Adversarial Networks(GANs),Vision Transformers(ViTs),and hybrid models,are discussed in detail,including their advantages and domain-specific adaptations.Advanced learning paradigms such as semi-supervised learning,selfsupervised learning,and few-shot learning are explored for their potential to mitigate data annotation challenges in clinical datasets.This review further categorizes major tasks in medical image analysis,elaborating on how DL techniques have enabled precise tumor segmentation,lesion detection,modality fusion,super-resolution,and robust classification across diverse clinical settings.Emphasis is placed on applications in oncology,cardiology,neurology,and infectious diseases,including COVID-19.Challenges such as data scarcity,label imbalance,model generalizability,interpretability,and integration into clinical workflows are critically examined.Ethical considerations,explainable AI(XAI),federated learning,and regulatory compliance are discussed as essential components of real-world deployment.Benchmark datasets,evaluation metrics,and comparative performance analyses are presented to support future research.The article concludes with a forward-looking perspective on the role of foundation models,multimodal learning,edge AI,and bio-inspired computing in the future of medical imaging.Overall,this review serves as a valuable resource for researchers,clinicians,and developers aiming to harness deep learning for intelligent,efficient,and clinically viable medical image analysis.
基金Supported by National Natural Science Foundation of China(Grant Nos.51175007,51075023)
文摘The condition monitoring and fault diagnosis of rolling element bearings are particularly crucial in rotating mechanical applications in industry. A bearing fault signal contains information not only about fault condition and fault type but also the severity of the fault. This means fault severity quantitative analysis is one of most active and valid ways to realize proper maintenance decision. Aiming at the deficiency of the research in bearing single point pitting fault quantitative diagnosis, a new back-propagation neural network method based on wavelet packet decomposition coefficient entropy is proposed. The three levels of wavelet packet coefficient entropy（WPCE） is introduced as a characteristic input vector to the BPNN. Compared with the wavelet packet decomposition energy ratio input vector, WPCE shows more sensitive in distinguishing from the different fault severity degree of the measured signal. The engineering application results show that the quantitative trend fault diagnosis is realized in the different fault degree of the single point bearing pitting fault. The breakthrough attempt from quantitative to qualitative on the pattern recognition of rolling element bearings fault diagnosis is realized.
基金supported by the“Regional Innovation System&Education(RISE)”through the Seoul RISE Center,funded by the Ministry of Education(MOE)and the Seoul Metropolitan Government(2025-RISE-01-018-05)supported by Quad Miners Corp。
文摘The dynamic,heterogeneous nature of Edge computing in the Internet of Things(Edge-IoT)and Industrial IoT(IIoT)networks brings unique and evolving cybersecurity challenges.This study maps cyber threats in Edge-IoT/IIoT environments to the Adversarial Tactics,Techniques,and Common Knowledge(ATT&CK)framework by MITRE and introduces a lightweight,data-driven scoring model that enables rapid identification and prioritization of attacks.Inspired by the Factor Analysis of Information Risk model,our proposed scoring model integrates four key metrics:Common Vulnerability Scoring System(CVSS)-based severity scoring,Cyber Kill Chain–based difficulty estimation,Deep Neural Networks-driven detection scoring,and frequency analysis based on dataset prevalence.By aggregating these indicators,the model generates comprehensive risk profiles,facilitating actionable prioritization of threats.Robustness and stability of the scoring model are validated through non-parametric correlation analysis using Spearman’s and Kendall’s rank correlation coefficients,demonstrating consistent performance across diverse scenarios.The approach culminates in a prioritized attack ranking that provides actionable guidance for risk mitigation and resource allocation in Edge-IoT/IIoT security operations.By leveraging real-world data to align MITRE ATT&CK techniques with CVSS metrics,the framework offers a standardized and practically applicable solution for consistent threat assessment in operational settings.The proposed lightweight scoring model delivers rapid and reliable results under dynamic cyber conditions,facilitating timely identification of attack scenarios and prioritization of response strategies.Our systematic integration of established taxonomies with data-driven indicators strengthens practical risk management and supports strategic planning in next-generation IoT deployments.Ultimately,this work advances adaptive threat modeling for Edge/IIoT ecosystems and establishes a robust foundation for evidence-based prioritization in emerging cyber-physical infrastructures.
基金This study was supported by the Independent Innovation Science Foundation Project of National University of Defense Technology,China(No.22-ZZCX-083).
文摘Qualitative spacecraft pursuit-evasion problem which focuses on feasibility is rarely studied because of high-dimensional dynamics,intractable terminal constraints and heavy computational cost.In this paper,A physics-informed framework is proposed for the problem,providing an intuitive method for spacecraft threat relationship determination,situation assessment,mission feasibility analysis and orbital game rules summarization.For the first time,situation adjustment suggestions can be provided for the weak player in orbital game.First,a dimension-reduction dynamics is derived in the line-of-sight rotation coordinate system and the qualitative model is determined,reducing complexity and avoiding the difficulty of target set presentation caused by individual modeling.Second,the Backwards Reachable Set(BRS)of the target set is used for state space partition and capture zone presentation.Reverse-time analysis can eliminate the influence of changeable initial state and enable the proposed framework to analyze plural situations simultaneously.Third,a time-dependent Hamilton-Jacobi-Isaacs(HJI)Partial Differential Equation(PDE)is established to describe BRS evolution driven by dimension-reduction dynamics,based on level set method.Then,Physics-Informed Neural Networks(PINNs)are extended to HJI PDE final value problem,supporting orbital game rules summarization through capture zone evolution analysis.Finally,numerical results demonstrate the feasibility and efficiency of the proposed framework.
