Background: Over the last decades, many forest simulators have been developed for the forests of individual European countries. The underlying growth models are usually based on national datasets of varying size, obta...Background: Over the last decades, many forest simulators have been developed for the forests of individual European countries. The underlying growth models are usually based on national datasets of varying size, obtained from National Forest Inventories or from long-term research plots. Many of these models include country-and location-specific predictors, such as site quality indices that may aggregate climate, soil properties and topography effects. Consequently, it is not sensible to compare such models among countries, and it is often impossible to apply models outside the region or country they were developed for. However, there is a clear need for more generically applicable but still locally accurate and climate sensitive simulators at the European scale, which requires the development of models that are applicable across the European continent. The purpose of this study is to develop tree diameter increment models that are applicable at the European scale, but still locally accurate. We compiled and used a dataset of diameter increment observations of over 2.3 million trees from 10 National Forest Inventories in Europe and a set of 99 potential explanatory variables covering forest structure, weather, climate, soil and nutrient deposition.Results: Diameter increment models are presented for 20 species/species groups. Selection of explanatory variables was done using a combination of forward and backward selection methods. The explained variance ranged from10% to 53% depending on the species. Variables related to forest structure(basal area of the stand and relative size of the tree) contributed most to the explained variance, but environmental variables were important to account for spatial patterns. The type of environmental variables included differed greatly among species.Conclusions: The presented diameter increment models are the first of their kind that are applicable at the European scale. This is an important step towards the development of a new generation of forest development simulators that can be applied at the European scale, but that are sensitive to variations in growing conditions and applicable to a wider range of management systems than before. This allows European scale but detailed analyses concerning topics like CO2 sequestration, wood mobilisation, long term impact of management, etc.展开更多
Three groups of dynamic triaxial tests were performed for saturated Nanjing fine sand subjected to uniform cyclic loading. The tested curves of the excess pore water pressure （EPWP） ratio variation with the ratio of...Three groups of dynamic triaxial tests were performed for saturated Nanjing fine sand subjected to uniform cyclic loading. The tested curves of the excess pore water pressure （EPWP） ratio variation with the ratio of the number of cycles are provided. The concept of the EPWP increment ratio is introduced and two new concepts of the effective dynamic shear stress ratio and the log decrement of effective stress are defined. It is found that the development of the EPWP increment ratio can be divided into three stages： descending, stable and ascending. Furthermore, at the stable and ascending stages, a satisfactory linear relationship is obtained between the accumulative EPWP increment ratio and natural logarithm of the effective dynamic shear stress ratio. Accordingly, the EPWP increment ratio at the number of cycles N has been deduced that is proportional to the log decrement of effective stress at the cycle number N-l, but is independent of the cyclic stress amplitude. Based on the analysis, a new EPWP increment model for saturated Nanjing fine sand is developed from tested data fitting, which provides a better prediction of the curves of EPWP generation, the number of cycles required for initial liquefaction and the liquefaction resistance.展开更多
Fashion industry has a complex characteristic for it spans the first, second, and third industries. In addition, the characteristic of creative industry has high value-added for its knowledge outputting, which makes t...Fashion industry has a complex characteristic for it spans the first, second, and third industries. In addition, the characteristic of creative industry has high value-added for its knowledge outputting, which makes the traditional value-added analysis based on supply chain not easy and good enough to interpret its industry value-added features. From the perspective of "products-knowledge" two-dimensional analysis,a fashion industry value chain increment model is built,by simulating the process of "product flow" and "information flow" value-added. The fashion industry value chain increment model provides an effective way for the enterprise strategy formulation and production strategy adjustment.展开更多
In this paper,an incremental contact model is developed for the elastic self-affine fractal rough surfaces under plane strain condition.The contact between a rough surface and a rigid plane is simplified by the accumu...In this paper,an incremental contact model is developed for the elastic self-affine fractal rough surfaces under plane strain condition.The contact between a rough surface and a rigid plane is simplified by the accumulation of identical line contacts with half-width given by the truncated area divided by the contact patch number at varying heights.Based on the contact stiffness of two-dimensional flat punch,the total stiffness of rough surface is estimated,and then the normal load is calculated by an incremental method.For various rough surfaces,the approximately linear load-area relationships predicted by the proposed model agree well with the results of finite element simulations.It is found that the real average contact pressure depends significantly on profile properties.展开更多
Fault diagnosis is an important measure to ensure the safety of production, and all kinds of fault diagnosis methods are of importance in actual production process. However, the complexity and uncertainty of productio...Fault diagnosis is an important measure to ensure the safety of production, and all kinds of fault diagnosis methods are of importance in actual production process. However, the complexity and uncertainty of production process often lead to the changes of data distribution and the emergence of new fault classes, and the number of the new fault classes is unpredictable. The reconstruction of the fault diagnosis model and the identification of new fault classes have become core issues under the circumstances. This paper presents a fault diagnosis method based on model transfer learning and the main contributions of the paper are as follows: 1) An incremental model transfer fault diagnosis method is proposed to reconstruct the new process diagnosis model. 2) Breaking the limit of existing method that the new process can only have one more class of faults than the old process, this method can identify M faults more in the new process with the thought of incremental learning. 3) The method offers a solution to a series of problems caused by the increase of fault classes. Experiments based on Tennessee-Eastman process and ore grinding classification process demonstrate the effectiveness and the feasibility of the method.展开更多
Three series of sandbox modeling experiments were performed to study the fault-increment pattern in extensional basins. Experimental results showed that the tectonic action mode of boundaries and the shape of major bo...Three series of sandbox modeling experiments were performed to study the fault-increment pattern in extensional basins. Experimental results showed that the tectonic action mode of boundaries and the shape of major boundary faults control the formation and evolution of faults in extensional basins. In the process of extensional deformation, the increase in the number and length of faults was episodic, and every ＇episode＇ experienced three periods, strain-accumulation period, quick fault-increment period and strain-adjustment period. The more complex the shape of the boundary fault, the higher the strain increment each ＇episode＇ experienced. Different extensional modes resulted in different fault-increment patterns. The horizontal detachment extensional mode has the ＇linear＇ style of fault-increment pattern, while the extensional mode controlled by a listric fault has the ＇stepwise＇ style of fault-increment pattern, and the extensional mode controlled by a ramp-flat boundary fault has the ＇stepwise-linear＇ style of fault-increment pattern. These fault-increment patterns given above could provide a theoretical method of fault interpretation and fracture prediction in extensional basins.展开更多
Warp yarns and weft yarns of plain woven fabric are the principal axes of mate- rial of fabric. They are orthogonal in their original con?guration, but are obliquely crisscross in deformed con?guration in general. I...Warp yarns and weft yarns of plain woven fabric are the principal axes of mate- rial of fabric. They are orthogonal in their original con?guration, but are obliquely crisscross in deformed con?guration in general. In this paper the expressions of incremental components of strain tensor are derived, the non-linear model of woven fabric is linearized physically and its geometric non-linearity survives. The convenience of determining the total deformation is shown by the choice of the coordinate system of the principal axes of the material, with the convergence of the incremental methods illustrated by examples. This incremental model furnishes a basis for numerical simulations of fabric draping and wrinkling based on the micro-mechanical model of fabric.展开更多
In this paper, we present a modular incremental statistical model for English full parsing. Unlike other full parsing approaches in which the analysis of the sentence is a uniform process, our model separates the full...In this paper, we present a modular incremental statistical model for English full parsing. Unlike other full parsing approaches in which the analysis of the sentence is a uniform process, our model separates the full parsing into shallow parsing and sentence skeleton parsing. In shallow parsing, we finish POS tagging, Base NP identification, prepositional phrase attachment and subordinate clause identification. In skeleton parsing, we use a layered feature-oriented statistical method. Modularity possesses the advantage of solving different problems in parsing with corresponding mechanisms. Feature-oriented rule is able to express the complex lingual phenomena at the key point if needed. Evaluated on Penn Treebank corpus, we obtained 89.2% precision and 89.8% recall.展开更多
Four-dimensional variational(4D-VAR) data assimilation method is a perfect data assimilation solution in theory, but the computational issue is quite difficult in operational implementation.The incremental 4D-VAR assi...Four-dimensional variational(4D-VAR) data assimilation method is a perfect data assimilation solution in theory, but the computational issue is quite difficult in operational implementation.The incremental 4D-VAR assimilation scheme is set up in order to reduce the computational cost. It is shown that the accuracy of the observations, the length of the assimilation window and the choice of the first guess have an important influence on the assimilation outcome through the contrast experiment. Compared with the standard 4D-VAR assimilation scheme, the incremental 4D-VAR assimilation scheme shows its advantage in the computation speed through an assimilation experiment.展开更多
Based on a multilevel linear mixed model approach,an individual diameter increment model was developed for fir plantation trees growing in Jiangxi Province.The data set used in this study came from long-term permanent...Based on a multilevel linear mixed model approach,an individual diameter increment model was developed for fir plantation trees growing in Jiangxi Province.The data set used in this study came from long-term permanent research plots.The database consists of total of 82 counties,365 plots, 5 416 trees and 16 248 observations.The paper chose mixed effects models instead of regression analysis approach because it allows for proper treatment of error terms and correlation in a repeated measures analysis framework.The model was defined as a mixed linear model with parameter random effect of plot,area or plot and area simultaneous.In addition the heteroscedasticity and correlation was taken into account.Mixed model calibration of diameter increment was carried out with the independent data using a different sample of complementary observations.The result showed that the total stand basal area,the diameter of target trees,the ratio of basal area of larger trees to target tree diameter,and altitude were found to be significant predictors.Both the fitting model and the calibrated model mean a substantial improvement compared with the classical approach widely used in forest management.After taking into account reasonable variance function of heteroscedasticity and correlation,the model shows better of goodness of fit than only taking into account parameter random effects.This type of modeling methodology shows flexible,precise and accurate.展开更多
社交网络中,节点间存在多种关系类型,节点数量会随着时间的推移而变化,这种异质性和动态性给链路预测任务带来极大的挑战。因此,本文提出一种基于增量学习的社交网络链路预测方法(incremental learning social networks link prediction...社交网络中,节点间存在多种关系类型,节点数量会随着时间的推移而变化,这种异质性和动态性给链路预测任务带来极大的挑战。因此,本文提出一种基于增量学习的社交网络链路预测方法(incremental learning social networks link prediction,IL-SNLP)。通过对网络进行分层,使每一层网络只包含一种关系类型,以更好地获取节点在每种关系类型下的语义信息;针对网络的动态性,利用时序随机游走捕获社交网络中的局部结构信息和时序信息;针对增量数据,采用增量式更新随机游走策略对历史随机游走序列进行更新。通过增量式skip-gram模型从随机游走序列中提取新出现节点的特征,并进一步更新历史节点的特征;针对网络的异质性,采用概率模型提取不同关系类型之间的因果关系关联程度,并将其作用于每一层的节点特征,以改善不同关系层下节点特征表现能力;利用多层感知机构建节点相互感知器,挖掘节点间建立连接时的相互贡献,实现更高的链路预测准确率。实验结果表明,在3个真实的社交网络数据集上,IL-SNLP方法的ROC曲线下的面积(AUC)和F1分数比基线方法分别提高了10.08%~67.60%和1.76%~64.67%,提升了预测性能;对于增量数据,只需要少次迭代就能保持预测模型的性能,提高了模型训练的速度;与未采用增量学习技术的IL-SNLP−方法相比,IL-SNLP方法在时间效率上提升了30.78%~257.58%,显著缩短了模型的运行时长。展开更多
为了解决真实Web应用攻击数据数量小、差异性大和攻击载荷多样化导致大模型训练效果差的问题,提出一种基于联邦大模型的网络攻击检测方法(Intrusion Detection methods based on Federal Large Language Model,FLLLMID).首先,提出一种...为了解决真实Web应用攻击数据数量小、差异性大和攻击载荷多样化导致大模型训练效果差的问题,提出一种基于联邦大模型的网络攻击检测方法(Intrusion Detection methods based on Federal Large Language Model,FLLLMID).首先,提出一种面向大模型微调的联邦学习网络,服务器对客户端本地大模型通过增量数据训练产生的参数,进行增量聚合的方式,提高联邦学习中大模型的参数聚合效率以及避免网络流量数据暴露的问题;其次,基于大模型对代码的理解能力,提出面向应用层数据的攻击检测模型(CodeBERT-LSTM),通过对应用层数据报文进行分析,使用CodeBERT模型对有效字段进行向量编码后,结合长短期记忆网络(Long Short-Term Memory,LSTM)进行分类,实现对Web应用高效的攻击检测任务;最后,实验结果表明,FL-LLMID方法在面向应用层数据的攻击检测任务中准确率达到99.63%,与传统联邦学习相比,增量式学习的效率提升了12个百分点.展开更多
打孔盗油事件不但给国家造成巨大经济损失,还可能危害国家能源安全、生态安全和公共安全,对打孔盗油须防患于未然,以避免危害结果的发生。通过深入分析打孔盗油嫌疑车辆的行为特征,引入多维度增量式DBSCAN算法(increment Density-Based ...打孔盗油事件不但给国家造成巨大经济损失,还可能危害国家能源安全、生态安全和公共安全,对打孔盗油须防患于未然,以避免危害结果的发生。通过深入分析打孔盗油嫌疑车辆的行为特征,引入多维度增量式DBSCAN算法(increment Density-Based Spatial Clustering of Applications with Noise,IncDBSCAN),挖掘公安视频监控车辆抓拍数据中的潜在规律,可有效识别在管道保护区内活动的涉嫌盗油异常车辆。在与传统的支持向量机(Support Vector Machine,SVM)、基于密度带有噪声的空间聚类算法(Density-Based Spatial Clustering of Applications with Noise,DBSCAN)、K均值聚类算法(K-Means Clustering Algorithm,K-Means)的对比试验中,多维度IncDBSCAN模型具有更好的检测效果,其精确率为85%,召回率为82%,F1值为83.4%,均优于其他模型。该方法为输油管道打孔盗油视频智能预警提供了一种新的思路和手段。展开更多
基金funded by the SIMWOOD project(Grant Agreement No.613762)of the EU H2020 Programmefacilitated by the Alter For project(Grant Agreement No.676754)+3 种基金the VERIFY project(Grant Agreement No.776810)Co-funding was received from the topsector Agri&Food under No.AF-EU-15002The Dutch National Forest Inventory is funded by the Ministry of Economic AffairsThe regional forest inventory in Piemonte was produced with the support of EU structural funds
文摘Background: Over the last decades, many forest simulators have been developed for the forests of individual European countries. The underlying growth models are usually based on national datasets of varying size, obtained from National Forest Inventories or from long-term research plots. Many of these models include country-and location-specific predictors, such as site quality indices that may aggregate climate, soil properties and topography effects. Consequently, it is not sensible to compare such models among countries, and it is often impossible to apply models outside the region or country they were developed for. However, there is a clear need for more generically applicable but still locally accurate and climate sensitive simulators at the European scale, which requires the development of models that are applicable across the European continent. The purpose of this study is to develop tree diameter increment models that are applicable at the European scale, but still locally accurate. We compiled and used a dataset of diameter increment observations of over 2.3 million trees from 10 National Forest Inventories in Europe and a set of 99 potential explanatory variables covering forest structure, weather, climate, soil and nutrient deposition.Results: Diameter increment models are presented for 20 species/species groups. Selection of explanatory variables was done using a combination of forward and backward selection methods. The explained variance ranged from10% to 53% depending on the species. Variables related to forest structure(basal area of the stand and relative size of the tree) contributed most to the explained variance, but environmental variables were important to account for spatial patterns. The type of environmental variables included differed greatly among species.Conclusions: The presented diameter increment models are the first of their kind that are applicable at the European scale. This is an important step towards the development of a new generation of forest development simulators that can be applied at the European scale, but that are sensitive to variations in growing conditions and applicable to a wider range of management systems than before. This allows European scale but detailed analyses concerning topics like CO2 sequestration, wood mobilisation, long term impact of management, etc.
基金Key Research Project of National Natural Science Foundation of China Under Grant No.90715018National Basic Research Program of China Under Grant No.2007CB714200the Special Fund for the Commonweal Industry of China Under Grant No.200808022
文摘Three groups of dynamic triaxial tests were performed for saturated Nanjing fine sand subjected to uniform cyclic loading. The tested curves of the excess pore water pressure （EPWP） ratio variation with the ratio of the number of cycles are provided. The concept of the EPWP increment ratio is introduced and two new concepts of the effective dynamic shear stress ratio and the log decrement of effective stress are defined. It is found that the development of the EPWP increment ratio can be divided into three stages： descending, stable and ascending. Furthermore, at the stable and ascending stages, a satisfactory linear relationship is obtained between the accumulative EPWP increment ratio and natural logarithm of the effective dynamic shear stress ratio. Accordingly, the EPWP increment ratio at the number of cycles N has been deduced that is proportional to the log decrement of effective stress at the cycle number N-l, but is independent of the cyclic stress amplitude. Based on the analysis, a new EPWP increment model for saturated Nanjing fine sand is developed from tested data fitting, which provides a better prediction of the curves of EPWP generation, the number of cycles required for initial liquefaction and the liquefaction resistance.
基金Shanghai University Young Teachers Training Program,China(No.KY01X0322016010)
文摘Fashion industry has a complex characteristic for it spans the first, second, and third industries. In addition, the characteristic of creative industry has high value-added for its knowledge outputting, which makes the traditional value-added analysis based on supply chain not easy and good enough to interpret its industry value-added features. From the perspective of "products-knowledge" two-dimensional analysis,a fashion industry value chain increment model is built,by simulating the process of "product flow" and "information flow" value-added. The fashion industry value chain increment model provides an effective way for the enterprise strategy formulation and production strategy adjustment.
基金supported by the National Natural Science Foundation of China(Grant Nos.12372100,12302126,and 12302141)the China Postdoctoral Science Foundation(Grant No.2023M732799)+1 种基金the Fundamental Research Funds for the Central Universities(Grant No.xzy012024020)Sihe Wang also thanks the support from the China Scholarship Council(CSC).
文摘In this paper,an incremental contact model is developed for the elastic self-affine fractal rough surfaces under plane strain condition.The contact between a rough surface and a rigid plane is simplified by the accumulation of identical line contacts with half-width given by the truncated area divided by the contact patch number at varying heights.Based on the contact stiffness of two-dimensional flat punch,the total stiffness of rough surface is estimated,and then the normal load is calculated by an incremental method.For various rough surfaces,the approximately linear load-area relationships predicted by the proposed model agree well with the results of finite element simulations.It is found that the real average contact pressure depends significantly on profile properties.
文摘Fault diagnosis is an important measure to ensure the safety of production, and all kinds of fault diagnosis methods are of importance in actual production process. However, the complexity and uncertainty of production process often lead to the changes of data distribution and the emergence of new fault classes, and the number of the new fault classes is unpredictable. The reconstruction of the fault diagnosis model and the identification of new fault classes have become core issues under the circumstances. This paper presents a fault diagnosis method based on model transfer learning and the main contributions of the paper are as follows: 1) An incremental model transfer fault diagnosis method is proposed to reconstruct the new process diagnosis model. 2) Breaking the limit of existing method that the new process can only have one more class of faults than the old process, this method can identify M faults more in the new process with the thought of incremental learning. 3) The method offers a solution to a series of problems caused by the increase of fault classes. Experiments based on Tennessee-Eastman process and ore grinding classification process demonstrate the effectiveness and the feasibility of the method.
文摘Three series of sandbox modeling experiments were performed to study the fault-increment pattern in extensional basins. Experimental results showed that the tectonic action mode of boundaries and the shape of major boundary faults control the formation and evolution of faults in extensional basins. In the process of extensional deformation, the increase in the number and length of faults was episodic, and every ＇episode＇ experienced three periods, strain-accumulation period, quick fault-increment period and strain-adjustment period. The more complex the shape of the boundary fault, the higher the strain increment each ＇episode＇ experienced. Different extensional modes resulted in different fault-increment patterns. The horizontal detachment extensional mode has the ＇linear＇ style of fault-increment pattern, while the extensional mode controlled by a listric fault has the ＇stepwise＇ style of fault-increment pattern, and the extensional mode controlled by a ramp-flat boundary fault has the ＇stepwise-linear＇ style of fault-increment pattern. These fault-increment patterns given above could provide a theoretical method of fault interpretation and fracture prediction in extensional basins.
基金Project supported by the National Natural Science Foundation of China (No. 10272079).
文摘Warp yarns and weft yarns of plain woven fabric are the principal axes of mate- rial of fabric. They are orthogonal in their original con?guration, but are obliquely crisscross in deformed con?guration in general. In this paper the expressions of incremental components of strain tensor are derived, the non-linear model of woven fabric is linearized physically and its geometric non-linearity survives. The convenience of determining the total deformation is shown by the choice of the coordinate system of the principal axes of the material, with the convergence of the incremental methods illustrated by examples. This incremental model furnishes a basis for numerical simulations of fabric draping and wrinkling based on the micro-mechanical model of fabric.
文摘In this paper, we present a modular incremental statistical model for English full parsing. Unlike other full parsing approaches in which the analysis of the sentence is a uniform process, our model separates the full parsing into shallow parsing and sentence skeleton parsing. In shallow parsing, we finish POS tagging, Base NP identification, prepositional phrase attachment and subordinate clause identification. In skeleton parsing, we use a layered feature-oriented statistical method. Modularity possesses the advantage of solving different problems in parsing with corresponding mechanisms. Feature-oriented rule is able to express the complex lingual phenomena at the key point if needed. Evaluated on Penn Treebank corpus, we obtained 89.2% precision and 89.8% recall.
基金the National Basic Research Program of China under Natural contract Nos 2007CB816001 and 2006CB400603Natinal Natural Science Foundation of China under contract Nos 40346027 and 40676008the China"908"-Project under Grant No.908-02-01-03 and 908-IC-I-13
文摘Four-dimensional variational(4D-VAR) data assimilation method is a perfect data assimilation solution in theory, but the computational issue is quite difficult in operational implementation.The incremental 4D-VAR assimilation scheme is set up in order to reduce the computational cost. It is shown that the accuracy of the observations, the length of the assimilation window and the choice of the first guess have an important influence on the assimilation outcome through the contrast experiment. Compared with the standard 4D-VAR assimilation scheme, the incremental 4D-VAR assimilation scheme shows its advantage in the computation speed through an assimilation experiment.
文摘Based on a multilevel linear mixed model approach,an individual diameter increment model was developed for fir plantation trees growing in Jiangxi Province.The data set used in this study came from long-term permanent research plots.The database consists of total of 82 counties,365 plots, 5 416 trees and 16 248 observations.The paper chose mixed effects models instead of regression analysis approach because it allows for proper treatment of error terms and correlation in a repeated measures analysis framework.The model was defined as a mixed linear model with parameter random effect of plot,area or plot and area simultaneous.In addition the heteroscedasticity and correlation was taken into account.Mixed model calibration of diameter increment was carried out with the independent data using a different sample of complementary observations.The result showed that the total stand basal area,the diameter of target trees,the ratio of basal area of larger trees to target tree diameter,and altitude were found to be significant predictors.Both the fitting model and the calibrated model mean a substantial improvement compared with the classical approach widely used in forest management.After taking into account reasonable variance function of heteroscedasticity and correlation,the model shows better of goodness of fit than only taking into account parameter random effects.This type of modeling methodology shows flexible,precise and accurate.
文摘社交网络中,节点间存在多种关系类型,节点数量会随着时间的推移而变化,这种异质性和动态性给链路预测任务带来极大的挑战。因此,本文提出一种基于增量学习的社交网络链路预测方法(incremental learning social networks link prediction,IL-SNLP)。通过对网络进行分层,使每一层网络只包含一种关系类型,以更好地获取节点在每种关系类型下的语义信息;针对网络的动态性,利用时序随机游走捕获社交网络中的局部结构信息和时序信息;针对增量数据,采用增量式更新随机游走策略对历史随机游走序列进行更新。通过增量式skip-gram模型从随机游走序列中提取新出现节点的特征,并进一步更新历史节点的特征;针对网络的异质性,采用概率模型提取不同关系类型之间的因果关系关联程度,并将其作用于每一层的节点特征,以改善不同关系层下节点特征表现能力;利用多层感知机构建节点相互感知器,挖掘节点间建立连接时的相互贡献,实现更高的链路预测准确率。实验结果表明,在3个真实的社交网络数据集上,IL-SNLP方法的ROC曲线下的面积(AUC)和F1分数比基线方法分别提高了10.08%~67.60%和1.76%~64.67%,提升了预测性能;对于增量数据,只需要少次迭代就能保持预测模型的性能,提高了模型训练的速度;与未采用增量学习技术的IL-SNLP−方法相比,IL-SNLP方法在时间效率上提升了30.78%~257.58%,显著缩短了模型的运行时长。
文摘为了解决真实Web应用攻击数据数量小、差异性大和攻击载荷多样化导致大模型训练效果差的问题,提出一种基于联邦大模型的网络攻击检测方法(Intrusion Detection methods based on Federal Large Language Model,FLLLMID).首先,提出一种面向大模型微调的联邦学习网络,服务器对客户端本地大模型通过增量数据训练产生的参数,进行增量聚合的方式,提高联邦学习中大模型的参数聚合效率以及避免网络流量数据暴露的问题;其次,基于大模型对代码的理解能力,提出面向应用层数据的攻击检测模型(CodeBERT-LSTM),通过对应用层数据报文进行分析,使用CodeBERT模型对有效字段进行向量编码后,结合长短期记忆网络(Long Short-Term Memory,LSTM)进行分类,实现对Web应用高效的攻击检测任务;最后,实验结果表明,FL-LLMID方法在面向应用层数据的攻击检测任务中准确率达到99.63%,与传统联邦学习相比,增量式学习的效率提升了12个百分点.
文摘打孔盗油事件不但给国家造成巨大经济损失,还可能危害国家能源安全、生态安全和公共安全,对打孔盗油须防患于未然,以避免危害结果的发生。通过深入分析打孔盗油嫌疑车辆的行为特征,引入多维度增量式DBSCAN算法(increment Density-Based Spatial Clustering of Applications with Noise,IncDBSCAN),挖掘公安视频监控车辆抓拍数据中的潜在规律,可有效识别在管道保护区内活动的涉嫌盗油异常车辆。在与传统的支持向量机(Support Vector Machine,SVM)、基于密度带有噪声的空间聚类算法(Density-Based Spatial Clustering of Applications with Noise,DBSCAN)、K均值聚类算法(K-Means Clustering Algorithm,K-Means)的对比试验中,多维度IncDBSCAN模型具有更好的检测效果,其精确率为85%,召回率为82%,F1值为83.4%,均优于其他模型。该方法为输油管道打孔盗油视频智能预警提供了一种新的思路和手段。
