As a data-driven approach, Deep Learning(DL)-based fault diagnosis methods need to collect the relatively comprehensive data on machine fault types to achieve satisfactory performance. A mechanical system may include ...As a data-driven approach, Deep Learning(DL)-based fault diagnosis methods need to collect the relatively comprehensive data on machine fault types to achieve satisfactory performance. A mechanical system may include multiple submachines in the real-world. During condition monitoring of a mechanical system, fault data are distributed in a continuous flow of constantly generated information and new faults will inevitably occur in unconsidered submachines, which are also called machine increments. Therefore, adequately collecting fault data in advance is difficult. Limited by the characteristics of DL, training existing models directly with new fault data of new submachines leads to catastrophic forgetting of old tasks, while the cost of collecting all known data to retrain the models is excessively high. DL-based fault diagnosis methods cannot learn continually and adaptively in dynamic environments. A new Continual Learning Fault Diagnosis method(CLFD) is proposed in this paper to solve a series of fault diagnosis tasks with machine increments. The stability–plasticity dilemma is an intrinsic issue in continual learning. The core of CLFD is the proposed Dual-branch Adaptive Aggregation Residual Network(DAARN).Two types of residual blocks are created in each block layer of DAARN: steady and dynamic blocks. The stability–plasticity dilemma is solved by assigning them with adaptive aggregation weights to balance stability and plasticity, and a bi-level optimization program is used to optimize adaptive aggregation weights and model parameters. In addition, a feature-level knowledge distillation loss function is proposed to further overcome catastrophic forgetting. CLFD is then applied to the fault diagnosis case with machine increments. Results demonstrate that CLFD outperforms other continual learning methods and has satisfactory robustness.展开更多
The evolution of temperature field of the continual motion induction cladding and the depth of heat affected zone are studied in this study.A three-dimensional finite element model for the point type continual inducti...The evolution of temperature field of the continual motion induction cladding and the depth of heat affected zone are studied in this study.A three-dimensional finite element model for the point type continual induction cladding is established to investigate temperature distributions of fixed and motion induction cladding modes.The novel inductor is designed for cladding of curved surfaces.The modeling reliability is verified by the temperature measurements.The influence of process parameters on the maximum temperature and the generation and transfer of heat are studied.Quantitative calculation is performed to its melting rate to verify the temperature distribution and microstructures.The results show that a good metallurgical bond can be formed between the cladding layer and substrate.The melting rate gradually falls from the top of the cladding layer to the substrate,and the grain size in the substrate gradually rises.The heat affected zone is relatively small compared to integral heating.展开更多
Seismic microzonation for Almaty city for the first time use probabilistic approach and hazard is expressed in terms of not only macroseismic intensity,but also Peak Ground Acceleration(PGA).To account for the effects...Seismic microzonation for Almaty city for the first time use probabilistic approach and hazard is expressed in terms of not only macroseismic intensity,but also Peak Ground Acceleration(PGA).To account for the effects of local soil conditions,the continual approach proposed by A.S.Aleshin[1,2]was used,in which soil coefficients are a function of the continuously changing seismic rigidity.Soil coefficients were calculated using the new data of geological and geophysical surveys and findings of previous geotechnical studies.The used approach made it possible to avoid using soil categories and a jump change in characteristics of soil conditions and seismic impact.The developed seismic microzonation maps are prepared for further introduction into the normative documents of the Republic of Kazakhstan.展开更多
Continual learning(CL)studies the problem of learning to accumulate knowledge over time from a stream of data.A crucial challenge is that neural networks suffer from performance degradation on previously seen data,kno...Continual learning(CL)studies the problem of learning to accumulate knowledge over time from a stream of data.A crucial challenge is that neural networks suffer from performance degradation on previously seen data,known as catastrophic forgetting,due to allowing parameter sharing.In this work,we consider a more practical online class-incremental CL setting,where the model learns new samples in an online manner and may continuously experience new classes.Moreover,prior knowledge is unavailable during training and evaluation.Existing works usually explore sample usages from a single dimension,which ignores a lot of valuable supervisory information.To better tackle the setting,we propose a novel replay-based CL method,which leverages multi-level representations produced by the intermediate process of training samples for replay and strengthens supervision to consolidate previous knowledge.Specifically,besides the previous raw samples,we store the corresponding logits and features in the memory.Furthermore,to imitate the prediction of the past model,we construct extra constraints by leveraging multi-level information stored in the memory.With the same number of samples for replay,our method can use more past knowledge to prevent interference.We conduct extensive evaluations on several popular CL datasets,and experiments show that our method consistently outperforms state-of-the-art methods with various sizes of episodic memory.We further provide a detailed analysis of these results and demonstrate that our method is more viable in practical scenarios.展开更多
In order to investigate and predict the material properties of curved surface AISI 1045 steel component during spot continual induction hardening(SCIH),a 3D model for curved surface workpieces which coupled electromag...In order to investigate and predict the material properties of curved surface AISI 1045 steel component during spot continual induction hardening(SCIH),a 3D model for curved surface workpieces which coupled electromagnetic,temperature and phase transformation fields was built by finite element software ANSYS.A small size inductor and magnetizer were used in this model,which can move along the top surface of workpiece flexibly.The effect of inductor moving velocity and workpiece radius on temperature field was analyzed and the heating delay phenomenon was found through comparing the simulated results.The temperature field results indicate that the heating delay phenomenon is more obvious under high inductor moving velocity condition.This trend becomes more obvious if the workpiece radius becomes larger.The predictions of microstructure and micro-hardness distribution were also carried out via this model.The predicted results show that the inductor moving velocity is the dominated factor for the distribution of 100% martensite region and phase transformation region.The influencing factor of workpiece radius on 100% martensite region and phase transformation region distribution is obvious under relatively high inductor moving velocity but inconspicuous under relatively low inductor moving velocity.展开更多
1|OVERVIEW.Machine learning(ML)has been increasingly used for tackling various diagnostic,therapeutic,and prognostic tasks owing to its capability to learn and reason without explicit programming[1].Most developed ML ...1|OVERVIEW.Machine learning(ML)has been increasingly used for tackling various diagnostic,therapeutic,and prognostic tasks owing to its capability to learn and reason without explicit programming[1].Most developed ML models have had their accuracy proven through internal validation using retrospective data.However,external validation using retrospective data,continual monitoring using prospective data,and randomized controlled trials(RCTs)using prospective data are important for the translation of ML models into real-world clinical practice[2].展开更多
In this paper, various forms of functional on blending energy principles of composite laminated plates are gir en, which guarantee satisfied continual conditions of displacements and stress between layers, and then th...In this paper, various forms of functional on blending energy principles of composite laminated plates are gir en, which guarantee satisfied continual conditions of displacements and stress between layers, and then the reliability of the functional are proved by the computing example.展开更多
Climate change poses significant challenges to agricultural management,particularly in adapting to extreme weather conditions that impact agricultural production.Existing works with traditional Reinforcement Learning(...Climate change poses significant challenges to agricultural management,particularly in adapting to extreme weather conditions that impact agricultural production.Existing works with traditional Reinforcement Learning(RL)methods often falter under such extreme conditions.To address this challenge,our study introduces a novel approach by integrating Continual Learning(CL)with RL to form Continual Reinforcement Learning(CRL),enhancing the adaptability of agricultural management strategies.Leveraging the Gym-DSSAT simulation environment,our research enables RL agents to learn optimal fertilization strategies based on variable weather conditions.By incorporating CL algorithms,such as Elastic Weight Consolidation(EWC),with established RL techniques like Deep Q-Networks(DQN),we developed a framework in which agents can learn and retain knowledge across diverse weather scenarios.The CRL approach was tested under climate variability to assess the robustness and adaptability of the induced policies,particularly under extreme weather events like severe droughts.Our results showed that continually learned policies exhibited superior adaptability and performance compared to optimal policies learned through the conventional RL methods,especially in challenging conditions of reduced rainfall and increased temperatures.This pioneering work,which combines CL with RL to generate adaptive policies for agricultural management,is expected to make significant advancements in precision agriculture in the era of climate change.展开更多
For the famous Feigenbaum's equations, in this paper, we established its constructive theorem of the peak-unimodal, then we found out other paths to explore the peak-unimodal solutions. For example, we proceed on ...For the famous Feigenbaum's equations, in this paper, we established its constructive theorem of the peak-unimodal, then we found out other paths to explore the peak-unimodal solutions. For example, we proceed on the direction to try the non-symmetrical continuous peak-unimodal solutions and C1 solutions.展开更多
Basic concepts of seismic zonation in Russia are the degree of intensity and soil categories that correspond to discrete structure in the ratio “seismic impact-ground reaction”. Meanwhile, the parameters of seismic ...Basic concepts of seismic zonation in Russia are the degree of intensity and soil categories that correspond to discrete structure in the ratio “seismic impact-ground reaction”. Meanwhile, the parameters of seismic effects, and the parameters of soil properties are continuous in the space. The report expounds the basic theory, adequately representing the above mentioned continuality. Thus, many the concepts of seismic zonation, used now, become either more correct, or unnecessary.展开更多
Introduction 2013 saw the sixth anniversary of the formulation of Regulations of the People＇s Republic of China on Disclosure of Government Information （hereinafter referred to as the Regulations） and the fifth ann...Introduction 2013 saw the sixth anniversary of the formulation of Regulations of the People＇s Republic of China on Disclosure of Government Information （hereinafter referred to as the Regulations） and the fifth anniversary of their being put into effect. As the Regulations were stipulated and put into force,展开更多
A method to reparametrize G retional curve to obtain a C^1 curve is given. A practical G^1 continual connective between adjacent NURUS patches along common guadratic boundary curve is presented in this paper, and a s...A method to reparametrize G retional curve to obtain a C^1 curve is given. A practical G^1 continual connective between adjacent NURUS patches along common guadratic boundary curve is presented in this paper, and a specific algorithm for control points and weights of NURBS patches is discussed.展开更多
Existing examinations in postgraduate education and continuing medical education (CME) are not perfect. Modern assessment does not reflect disadvantages of older responders, for whom more time for reply is needed. Spe...Existing examinations in postgraduate education and continuing medical education (CME) are not perfect. Modern assessment does not reflect disadvantages of older responders, for whom more time for reply is needed. Specialists with wide clinical experience may choose more than one correct answer in alternative questions. Reduced ability to remember in older people restricts examination without additional sources of information. We offer an individualised system for testing doctors. It provides personalised choice of examination questions using multiple choice questions with weight characteristics and absence of distractors, interactive cooperation in case of negative answers and the final decision of an expert in relation to the person tested. A special algorithm is proposed for typical questions that combines the advantages of known approaches to testing. The questioning system is complex for the creators of tests, but is more convenient and objective than existing ones for medical doctors.展开更多
Continual learning(CL)has emerged as a crucial paradigm for learning from sequential data while retaining previous knowledge.Continual graph learning(CGL),characterized by dynamically evolving graphs from streaming da...Continual learning(CL)has emerged as a crucial paradigm for learning from sequential data while retaining previous knowledge.Continual graph learning(CGL),characterized by dynamically evolving graphs from streaming data,presents distinct challenges that demand efficient algorithms to prevent catastrophic forgetting.The first challenge stems from the interdependencies between different graph data,in which previous graphs infuence new data distributions.The second challenge is handling large graphs in an efficient manner.To address these challenges,we propose an eficient continual graph learner(E-CGL)in this paper.We address the interdependence issue by demonstrating the effectiveness of replay strategies and introducing a combined sampling approach that considers both node importance and diversity.To improve efficiency,E-CGL leverages a simple yet effective multilayer perceptron(MLP)model that shares weights with a graph neural network(GNN)during training,thereby accelerating computation by circumventing the expensive message-passing process.Our method achieves state-ofthe-art results on four CGL datasets under two settings,while significantly lowering the catastrophic forgetting value to an average of-1.1%.Additionally,E-CGL achieves the training and inference speedup by an average of 15.83x and 4.89x,respectively,across four datasets.These results indicate that E-CGL not only effectively manages correlations between different graph data during continual training but also enhances efficiency in large-scale CGL.展开更多
The continuous investment into manpower resource, is the radical power toassure the sustainable development of enterprises. The enterprises both at home and abroad attachhigh importance to the continuous education of ...The continuous investment into manpower resource, is the radical power toassure the sustainable development of enterprises. The enterprises both at home and abroad attachhigh importance to the continuous education of their employees and consolidate training to inspiretheir employees. In order to face increasingly drastic global competition, the telecom enterprisesin our country should consolidate continuous education, make training plans to adapt to thelong-term development of the enterprises and establish the effective mechanism of encouragement ofcontinuous education.展开更多
Threshold public key encryption allows a set of servers to decrypt a ciphertext if a given threshold of authorized servers cooperate. In the setting of threshold public key encryption, we consider the question of how ...Threshold public key encryption allows a set of servers to decrypt a ciphertext if a given threshold of authorized servers cooperate. In the setting of threshold public key encryption, we consider the question of how to correctly decrypt a ciphertext where all servers continually leak information about their secret keys to an external attacker. Dodis et al. and Akavia et al. show two concrete schemes on how to store secrets on continually leaky servers. However, their construc- tions are only interactive between two servers. To achieve continual leakage security among more than two servers, we give the first threshold public key encryption scheme against adaptively chosen ciphertext attack in the continual leak- age model under three static assumptions. In our model, the servers update their keys individually and asynchronously, without any communication between two servers. Moreover, the update procedure is re-randomized and the randomness can leak as well.展开更多
Tramadol is a potent analgesic.However,the analgesia efficacy of tramadol,particularly its minimum effective dose(MED),is not clear.The aim of this study is to find MED of tramadol for postoperative analgesia in infan...Tramadol is a potent analgesic.However,the analgesia efficacy of tramadol,particularly its minimum effective dose(MED),is not clear.The aim of this study is to find MED of tramadol for postoperative analgesia in infants.The continual reassessment method(CRM)was performed to find MED.Infants undergoing surgeries were included in the 3 phases of this series.In each phase,24 participants were allocated a different tramadol dose.Pain intensity was measured by face,legs,activity,cry,consolability(FLACC)measurement at 3-hour intervals.Tramadol was considered ineffective if the FLACC score was higher than 4 in 10 at anytime.In phase 1,seven dose levels were used within the range 0.1-0.4 mg·kg^(-1)·h^(-1).Phase 1 was insufficient to identify the MED,and we increased the dose to 0.4-0.8 mg·kg^(-1)·h^(-1) in phase 2.Phase 2 was insufficient to identify the MED.In phase 3,local anesthetic wound infiltration was introduced,and the tramadol dose levels tested were the same as in phase 1.The successful analgesia probability of tramadol 0.4 mg·kg^(-1)·h^(-1) was 82.1%(95%CI,0.742-0.925)in phase 1.In phase 2,it was 84.7%(95%CI,0.789-0.991)with the dose 0.8 mg·kg^(-1)·h^(-1).Phase 1 and phase 2 were insufficient to identify the MED.In phase 3,the successful analgesia probability for dose 0.35 mg·kg^(-1)·h^(-1) was 96.7%(95%CI,0.853-0.997).We have demonstrated that tramadol provides insufficient analgesia for surgeries considered to cause moderate-to-severe postoperative pain in infants if used as the sole analgesic,and that local anesthetic wound infiltration enhances the efficacy of tramadol.展开更多
Modelling,predicting and analysing driver behaviours are essential to advanced driver assistance systems(ADAS)and the comprehensive understanding of complex driving scenarios.Recently,with the development of deep lear...Modelling,predicting and analysing driver behaviours are essential to advanced driver assistance systems(ADAS)and the comprehensive understanding of complex driving scenarios.Recently,with the development of deep learning(DL),numerous driver behaviour learning(DBL)methods have been proposed and applied in connected vehicles(CV)and intelligent transportation systems(ITS).This study provides a review of DBL,which mainly focuses on typical applications in CV and ITS.First,a comprehensive review of the state-of-the-art DBL is presented.Next,Given the constantly changing nature of real driving scenarios,most existing learning-based models may suffer from the so-called“catastrophic forgetting,”which refers to their inability to perform well in previously learned scenarios after acquiring new ones.As a solution to the aforementioned issue,this paper presents a framework for continual driver behaviour learning(CDBL)by leveraging continual learning technology.The proposed CDBL framework is demonstrated to outperform existing methods in behaviour prediction through a case study.Finally,future works,potential challenges and emerging trends in this area are highlighted.展开更多
Bayesian-based methods have emerged as an effective approach in continual learning(CL) to solve catastrophic forgetting. One prominent example is Variational Continual Learning(VCL), which demonstrates remarkable perf...Bayesian-based methods have emerged as an effective approach in continual learning(CL) to solve catastrophic forgetting. One prominent example is Variational Continual Learning(VCL), which demonstrates remarkable performance in task-incremental learning(task-IL).However, class-incremental learning(class-IL) is still challenging for VCL, and the reasons behind this limitation remain unclear. Relying on the sophisticated neural mechanisms, particularly the mechanism of memory consolidation during sleep, the human brain possesses inherent advantages for both task-IL and class-IL scenarios, which provides insight for a braininspired VCL. To identify the reasons for the inadequacy of VCL in class-IL, we first conduct a comprehensive theoretical analysis of VCL. On this basis, we propose a novel Bayesian framework named as Learning within Sleeping(Lw S) by leveraging the memory consolidation.By simulating the distribution integration and generalization observed during memory consolidation in sleep, Lw S achieves the idea of prior knowledge guiding posterior knowledge learning as in VCL. In addition, with emulating the process of memory reactivation of the brain,Lw S imposes a constraint on feature invariance to mitigate forgetting learned knowledge. Experimental results demonstrate that Lw S outperforms both Bayesian and non-Bayesian methods in task-IL and class-IL scenarios, which further indicates the effectiveness of incorporating brain mechanisms on designing novel approaches for CL.展开更多
基金supported by the National Natural Science Foundation of China(Nos.52272440,51875375)the China Postdoctoral Science Foundation Funded Project(No.2021M701503).
文摘As a data-driven approach, Deep Learning(DL)-based fault diagnosis methods need to collect the relatively comprehensive data on machine fault types to achieve satisfactory performance. A mechanical system may include multiple submachines in the real-world. During condition monitoring of a mechanical system, fault data are distributed in a continuous flow of constantly generated information and new faults will inevitably occur in unconsidered submachines, which are also called machine increments. Therefore, adequately collecting fault data in advance is difficult. Limited by the characteristics of DL, training existing models directly with new fault data of new submachines leads to catastrophic forgetting of old tasks, while the cost of collecting all known data to retrain the models is excessively high. DL-based fault diagnosis methods cannot learn continually and adaptively in dynamic environments. A new Continual Learning Fault Diagnosis method(CLFD) is proposed in this paper to solve a series of fault diagnosis tasks with machine increments. The stability–plasticity dilemma is an intrinsic issue in continual learning. The core of CLFD is the proposed Dual-branch Adaptive Aggregation Residual Network(DAARN).Two types of residual blocks are created in each block layer of DAARN: steady and dynamic blocks. The stability–plasticity dilemma is solved by assigning them with adaptive aggregation weights to balance stability and plasticity, and a bi-level optimization program is used to optimize adaptive aggregation weights and model parameters. In addition, a feature-level knowledge distillation loss function is proposed to further overcome catastrophic forgetting. CLFD is then applied to the fault diagnosis case with machine increments. Results demonstrate that CLFD outperforms other continual learning methods and has satisfactory robustness.
基金Project(51575415)supported by the National Natural Science Foundation of ChinaProject(2016CFA077)supported by the Natural Science Foundation of Hubei Province of ChinaProject(2018-YS-026)supported by the Excellent Dissertation Cultivation Funds of Wuhan University of Technology,China。
文摘The evolution of temperature field of the continual motion induction cladding and the depth of heat affected zone are studied in this study.A three-dimensional finite element model for the point type continual induction cladding is established to investigate temperature distributions of fixed and motion induction cladding modes.The novel inductor is designed for cladding of curved surfaces.The modeling reliability is verified by the temperature measurements.The influence of process parameters on the maximum temperature and the generation and transfer of heat are studied.Quantitative calculation is performed to its melting rate to verify the temperature distribution and microstructures.The results show that a good metallurgical bond can be formed between the cladding layer and substrate.The melting rate gradually falls from the top of the cladding layer to the substrate,and the grain size in the substrate gradually rises.The heat affected zone is relatively small compared to integral heating.
基金provided through the Ministry of Education and Sciencecarried out as a part of the project“Development of the Seismic Microzonation Map for the Territory of Almaty City on a New Methodical Base”(state registration No 0115RK02701)funded within the state funding.
文摘Seismic microzonation for Almaty city for the first time use probabilistic approach and hazard is expressed in terms of not only macroseismic intensity,but also Peak Ground Acceleration(PGA).To account for the effects of local soil conditions,the continual approach proposed by A.S.Aleshin[1,2]was used,in which soil coefficients are a function of the continuously changing seismic rigidity.Soil coefficients were calculated using the new data of geological and geophysical surveys and findings of previous geotechnical studies.The used approach made it possible to avoid using soil categories and a jump change in characteristics of soil conditions and seismic impact.The developed seismic microzonation maps are prepared for further introduction into the normative documents of the Republic of Kazakhstan.
基金supported in part by the National Natura Science Foundation of China(U2013602,61876181,51521003)the Nationa Key R&D Program of China(2020YFB13134)+2 种基金Shenzhen Science and Technology Research and Development Foundation(JCYJ20190813171009236)Beijing Nova Program of Science and Technology(Z191100001119043)the Youth Innovation Promotion Association,Chinese Academy of Sciences。
文摘Continual learning(CL)studies the problem of learning to accumulate knowledge over time from a stream of data.A crucial challenge is that neural networks suffer from performance degradation on previously seen data,known as catastrophic forgetting,due to allowing parameter sharing.In this work,we consider a more practical online class-incremental CL setting,where the model learns new samples in an online manner and may continuously experience new classes.Moreover,prior knowledge is unavailable during training and evaluation.Existing works usually explore sample usages from a single dimension,which ignores a lot of valuable supervisory information.To better tackle the setting,we propose a novel replay-based CL method,which leverages multi-level representations produced by the intermediate process of training samples for replay and strengthens supervision to consolidate previous knowledge.Specifically,besides the previous raw samples,we store the corresponding logits and features in the memory.Furthermore,to imitate the prediction of the past model,we construct extra constraints by leveraging multi-level information stored in the memory.With the same number of samples for replay,our method can use more past knowledge to prevent interference.We conduct extensive evaluations on several popular CL datasets,and experiments show that our method consistently outperforms state-of-the-art methods with various sizes of episodic memory.We further provide a detailed analysis of these results and demonstrate that our method is more viable in practical scenarios.
基金Project (51175392) supported by the National Natural Science Foundation of ChinaProject (2014BAA012) supported by the Key Project of Hubei Province Science & Technology Pillar Program,ChinaProjects (2012-IV-067,2013-VII-020) supported by the Fundamental Research Funds for the Central Universities of China
文摘In order to investigate and predict the material properties of curved surface AISI 1045 steel component during spot continual induction hardening(SCIH),a 3D model for curved surface workpieces which coupled electromagnetic,temperature and phase transformation fields was built by finite element software ANSYS.A small size inductor and magnetizer were used in this model,which can move along the top surface of workpiece flexibly.The effect of inductor moving velocity and workpiece radius on temperature field was analyzed and the heating delay phenomenon was found through comparing the simulated results.The temperature field results indicate that the heating delay phenomenon is more obvious under high inductor moving velocity condition.This trend becomes more obvious if the workpiece radius becomes larger.The predictions of microstructure and micro-hardness distribution were also carried out via this model.The predicted results show that the inductor moving velocity is the dominated factor for the distribution of 100% martensite region and phase transformation region.The influencing factor of workpiece radius on 100% martensite region and phase transformation region distribution is obvious under relatively high inductor moving velocity but inconspicuous under relatively low inductor moving velocity.
文摘1|OVERVIEW.Machine learning(ML)has been increasingly used for tackling various diagnostic,therapeutic,and prognostic tasks owing to its capability to learn and reason without explicit programming[1].Most developed ML models have had their accuracy proven through internal validation using retrospective data.However,external validation using retrospective data,continual monitoring using prospective data,and randomized controlled trials(RCTs)using prospective data are important for the translation of ML models into real-world clinical practice[2].
文摘In this paper, various forms of functional on blending energy principles of composite laminated plates are gir en, which guarantee satisfied continual conditions of displacements and stress between layers, and then the reliability of the functional are proved by the computing example.
基金support from the University of Iowa OVPR Interdisciplinary Scholars Program and the US Department of Education(ED#P116S210005)for this study.Kishlay Jha’s work is supported in part by the US National Institute of Health(NIH)and National Science Foundation(NSF)under grants R01LM014012-01A1 and ITE-2333740.
文摘Climate change poses significant challenges to agricultural management,particularly in adapting to extreme weather conditions that impact agricultural production.Existing works with traditional Reinforcement Learning(RL)methods often falter under such extreme conditions.To address this challenge,our study introduces a novel approach by integrating Continual Learning(CL)with RL to form Continual Reinforcement Learning(CRL),enhancing the adaptability of agricultural management strategies.Leveraging the Gym-DSSAT simulation environment,our research enables RL agents to learn optimal fertilization strategies based on variable weather conditions.By incorporating CL algorithms,such as Elastic Weight Consolidation(EWC),with established RL techniques like Deep Q-Networks(DQN),we developed a framework in which agents can learn and retain knowledge across diverse weather scenarios.The CRL approach was tested under climate variability to assess the robustness and adaptability of the induced policies,particularly under extreme weather events like severe droughts.Our results showed that continually learned policies exhibited superior adaptability and performance compared to optimal policies learned through the conventional RL methods,especially in challenging conditions of reduced rainfall and increased temperatures.This pioneering work,which combines CL with RL to generate adaptive policies for agricultural management,is expected to make significant advancements in precision agriculture in the era of climate change.
基金Projects supported by National Natural Science Foundation of China
文摘For the famous Feigenbaum's equations, in this paper, we established its constructive theorem of the peak-unimodal, then we found out other paths to explore the peak-unimodal solutions. For example, we proceed on the direction to try the non-symmetrical continuous peak-unimodal solutions and C1 solutions.
文摘Basic concepts of seismic zonation in Russia are the degree of intensity and soil categories that correspond to discrete structure in the ratio “seismic impact-ground reaction”. Meanwhile, the parameters of seismic effects, and the parameters of soil properties are continuous in the space. The report expounds the basic theory, adequately representing the above mentioned continuality. Thus, many the concepts of seismic zonation, used now, become either more correct, or unnecessary.
文摘Introduction 2013 saw the sixth anniversary of the formulation of Regulations of the People＇s Republic of China on Disclosure of Government Information （hereinafter referred to as the Regulations） and the fifth anniversary of their being put into effect. As the Regulations were stipulated and put into force,
文摘A method to reparametrize G retional curve to obtain a C^1 curve is given. A practical G^1 continual connective between adjacent NURUS patches along common guadratic boundary curve is presented in this paper, and a specific algorithm for control points and weights of NURBS patches is discussed.
文摘Existing examinations in postgraduate education and continuing medical education (CME) are not perfect. Modern assessment does not reflect disadvantages of older responders, for whom more time for reply is needed. Specialists with wide clinical experience may choose more than one correct answer in alternative questions. Reduced ability to remember in older people restricts examination without additional sources of information. We offer an individualised system for testing doctors. It provides personalised choice of examination questions using multiple choice questions with weight characteristics and absence of distractors, interactive cooperation in case of negative answers and the final decision of an expert in relation to the person tested. A special algorithm is proposed for typical questions that combines the advantages of known approaches to testing. The questioning system is complex for the creators of tests, but is more convenient and objective than existing ones for medical doctors.
基金Project supported by the National Natural Science Foundation of China(No.62272411)the Key R&D Projects in Zhejiang Province(Nos.2024C01106 and 2025C01030)the Zhejiang Natural Science Foundation(No.LRG25F020001)。
文摘Continual learning(CL)has emerged as a crucial paradigm for learning from sequential data while retaining previous knowledge.Continual graph learning(CGL),characterized by dynamically evolving graphs from streaming data,presents distinct challenges that demand efficient algorithms to prevent catastrophic forgetting.The first challenge stems from the interdependencies between different graph data,in which previous graphs infuence new data distributions.The second challenge is handling large graphs in an efficient manner.To address these challenges,we propose an eficient continual graph learner(E-CGL)in this paper.We address the interdependence issue by demonstrating the effectiveness of replay strategies and introducing a combined sampling approach that considers both node importance and diversity.To improve efficiency,E-CGL leverages a simple yet effective multilayer perceptron(MLP)model that shares weights with a graph neural network(GNN)during training,thereby accelerating computation by circumventing the expensive message-passing process.Our method achieves state-ofthe-art results on four CGL datasets under two settings,while significantly lowering the catastrophic forgetting value to an average of-1.1%.Additionally,E-CGL achieves the training and inference speedup by an average of 15.83x and 4.89x,respectively,across four datasets.These results indicate that E-CGL not only effectively manages correlations between different graph data during continual training but also enhances efficiency in large-scale CGL.
文摘The continuous investment into manpower resource, is the radical power toassure the sustainable development of enterprises. The enterprises both at home and abroad attachhigh importance to the continuous education of their employees and consolidate training to inspiretheir employees. In order to face increasingly drastic global competition, the telecom enterprisesin our country should consolidate continuous education, make training plans to adapt to thelong-term development of the enterprises and establish the effective mechanism of encouragement ofcontinuous education.
基金This work was supported by the Science and Technology on Communication Security Laboratory Foundation （9140C110301110C1103）, the Weaponry Equipment Pre-Research Foundation, the PLA General Armament Department （9140A04020311DZ02）, and the National Natural Science Foundation of China （61370203）.
文摘Threshold public key encryption allows a set of servers to decrypt a ciphertext if a given threshold of authorized servers cooperate. In the setting of threshold public key encryption, we consider the question of how to correctly decrypt a ciphertext where all servers continually leak information about their secret keys to an external attacker. Dodis et al. and Akavia et al. show two concrete schemes on how to store secrets on continually leaky servers. However, their construc- tions are only interactive between two servers. To achieve continual leakage security among more than two servers, we give the first threshold public key encryption scheme against adaptively chosen ciphertext attack in the continual leak- age model under three static assumptions. In our model, the servers update their keys individually and asynchronously, without any communication between two servers. Moreover, the update procedure is re-randomized and the randomness can leak as well.
文摘Tramadol is a potent analgesic.However,the analgesia efficacy of tramadol,particularly its minimum effective dose(MED),is not clear.The aim of this study is to find MED of tramadol for postoperative analgesia in infants.The continual reassessment method(CRM)was performed to find MED.Infants undergoing surgeries were included in the 3 phases of this series.In each phase,24 participants were allocated a different tramadol dose.Pain intensity was measured by face,legs,activity,cry,consolability(FLACC)measurement at 3-hour intervals.Tramadol was considered ineffective if the FLACC score was higher than 4 in 10 at anytime.In phase 1,seven dose levels were used within the range 0.1-0.4 mg·kg^(-1)·h^(-1).Phase 1 was insufficient to identify the MED,and we increased the dose to 0.4-0.8 mg·kg^(-1)·h^(-1) in phase 2.Phase 2 was insufficient to identify the MED.In phase 3,local anesthetic wound infiltration was introduced,and the tramadol dose levels tested were the same as in phase 1.The successful analgesia probability of tramadol 0.4 mg·kg^(-1)·h^(-1) was 82.1%(95%CI,0.742-0.925)in phase 1.In phase 2,it was 84.7%(95%CI,0.789-0.991)with the dose 0.8 mg·kg^(-1)·h^(-1).Phase 1 and phase 2 were insufficient to identify the MED.In phase 3,the successful analgesia probability for dose 0.35 mg·kg^(-1)·h^(-1) was 96.7%(95%CI,0.853-0.997).We have demonstrated that tramadol provides insufficient analgesia for surgeries considered to cause moderate-to-severe postoperative pain in infants if used as the sole analgesic,and that local anesthetic wound infiltration enhances the efficacy of tramadol.
基金Supported by the National Key Research and Development Program of China(No.2022ZD0115503).
文摘Modelling,predicting and analysing driver behaviours are essential to advanced driver assistance systems(ADAS)and the comprehensive understanding of complex driving scenarios.Recently,with the development of deep learning(DL),numerous driver behaviour learning(DBL)methods have been proposed and applied in connected vehicles(CV)and intelligent transportation systems(ITS).This study provides a review of DBL,which mainly focuses on typical applications in CV and ITS.First,a comprehensive review of the state-of-the-art DBL is presented.Next,Given the constantly changing nature of real driving scenarios,most existing learning-based models may suffer from the so-called“catastrophic forgetting,”which refers to their inability to perform well in previously learned scenarios after acquiring new ones.As a solution to the aforementioned issue,this paper presents a framework for continual driver behaviour learning(CDBL)by leveraging continual learning technology.The proposed CDBL framework is demonstrated to outperform existing methods in behaviour prediction through a case study.Finally,future works,potential challenges and emerging trends in this area are highlighted.
基金supported by the National Natural Science Foundation of China under Grant 62236001 and Grant 62325601
文摘Bayesian-based methods have emerged as an effective approach in continual learning(CL) to solve catastrophic forgetting. One prominent example is Variational Continual Learning(VCL), which demonstrates remarkable performance in task-incremental learning(task-IL).However, class-incremental learning(class-IL) is still challenging for VCL, and the reasons behind this limitation remain unclear. Relying on the sophisticated neural mechanisms, particularly the mechanism of memory consolidation during sleep, the human brain possesses inherent advantages for both task-IL and class-IL scenarios, which provides insight for a braininspired VCL. To identify the reasons for the inadequacy of VCL in class-IL, we first conduct a comprehensive theoretical analysis of VCL. On this basis, we propose a novel Bayesian framework named as Learning within Sleeping(Lw S) by leveraging the memory consolidation.By simulating the distribution integration and generalization observed during memory consolidation in sleep, Lw S achieves the idea of prior knowledge guiding posterior knowledge learning as in VCL. In addition, with emulating the process of memory reactivation of the brain,Lw S imposes a constraint on feature invariance to mitigate forgetting learned knowledge. Experimental results demonstrate that Lw S outperforms both Bayesian and non-Bayesian methods in task-IL and class-IL scenarios, which further indicates the effectiveness of incorporating brain mechanisms on designing novel approaches for CL.
