In this paper, based on new Lyapunov function, the asymptotic properties of the dynamic neural system with asymmetric connection weights are investigated. Since the dynamic neural system with asymmetric connection wei...In this paper, based on new Lyapunov function, the asymptotic properties of the dynamic neural system with asymmetric connection weights are investigated. Since the dynamic neural system with asymmetric connection weights is more general than that with symmetric ones, the new results are significant in both theory and applications. Specially the new result can cover the asymptotic stability results of linear systems as special cases.展开更多
This paper derives some sufficient conditions for exponential stability for the equilibrium point by dividing the state variables of the system according to the characters of the neural networks. The new conditions ar...This paper derives some sufficient conditions for exponential stability for the equilibrium point by dividing the state variables of the system according to the characters of the neural networks. The new conditions are described by some blocks of the interconnection matrix. An example is given to demonstrate the effectiveness of the proposed theory.展开更多
The bending strength of carbon?fiber/thermoplastic epoxy composites?(CF/TP-EP Compo.)?had?bi-linear increasewith increase of weight-average molecular weight (Mw) of matrix. The transition in the bending strength appea...The bending strength of carbon?fiber/thermoplastic epoxy composites?(CF/TP-EP Compo.)?had?bi-linear increasewith increase of weight-average molecular weight (Mw) of matrix. The transition in the bending strength appeared at around 55k of Mw (“k”?means 103). SEM observation of fractured surface of CF/TP-EP Compo. showed that the fracture mode changed from interfacial failure to fiber breakage dominated failure. The smooth surface of carbon fibers appeared at lower Mw than 55k while some resin remained on the fibers indicating good adhesion between carbon fiber and matrix at higher Mw than 55k. The interfacial shear strength between carbon fiber and matrix bi-linearly increased with an increase of Mw similarly to the bending strength of the composite, measured by the micro droplet test. The dynamic loss tanδ?of the matrix measured at 2?Hz also showed a bi-linear relationship with respect to Mw having a knee point at Mw = 55k. The connection probability of two cracks introduced on?each side of specimens also confirmed that the interfacial strength between carbon fiber and matrix is the key for the mechanical performance of CF/TP-EP Compo. in bending.展开更多
Natural connectivity has been recently proposed to efficiently characterize the structural robustness of complex networks.The natural connectivity,interpreted as the Helmholtz free energy of a network,can be derived f...Natural connectivity has been recently proposed to efficiently characterize the structural robustness of complex networks.The natural connectivity,interpreted as the Helmholtz free energy of a network,can be derived from the graph spectrum.We extend the concept of natural connectivity to weighted complex networks,in which the weight represents the number of multiple edges.We prove that the weighted natural connectivity changes monotonically when the weights are increased or decreased.We investigate the influence of weight on the network robustness within scenarios of weight changing and show that the weighted natural connectivity allows a precise quantitative analysis of the structural robustness for weighted complex networks.展开更多
针对容迟网络中节点存储空间有限从而出现大量丢包情况的问题,本文提出了一种适用于节点环境状态的拥塞控制管理策略(congestion control management strategy suitable for the state of the node environment,NEMS):该策略由节点间位...针对容迟网络中节点存储空间有限从而出现大量丢包情况的问题,本文提出了一种适用于节点环境状态的拥塞控制管理策略(congestion control management strategy suitable for the state of the node environment,NEMS):该策略由节点间位置差异相关的控制保留策略和节点自差异相关的丢包策略构成.首先根据节点剩余缓存空间的大小判断节点是处于忙碌状态(BS)还是崩溃状态(CS),当节点处于BS状态时,节点间位置差异相关的控制保留策略结合门限度和连接活跃值的概念分析是否留存新消息.节点自差异相关的丢包策略利用熵权法动态计算各个节点中不同消息属性权重,得到消息丢弃优先级.当拥塞发生,优先删除丢弃优先级大的消息.同时,引入ACK反馈机制,消除网络中已成功投递的冗余消息.仿真结果表明,与无缓存管理的路由策略相比,加入NEMS后使得消息成功投递率增加了约100.93%~200.67%,网络负载降低了约76.05%~81.42%.展开更多
脱机手写中文字符识别(handwritten Chinese character recognition,HCCR)在计算机视觉领域一直是一个巨大的挑战。相比传统方法,基于深度学习的网络通过训练大量数据在识别任务中取得了差异化的效果,但识别效果依旧处于发展过程中。基...脱机手写中文字符识别(handwritten Chinese character recognition,HCCR)在计算机视觉领域一直是一个巨大的挑战。相比传统方法,基于深度学习的网络通过训练大量数据在识别任务中取得了差异化的效果,但识别效果依旧处于发展过程中。基于此,结合DW卷积和残差连接设计了一种多分支残差模块,该模块通过DW卷积以较小的内存和参数量为代价来加深网络深度,增强网络的特征提取能力;再通过残差连接抑制网络梯度问题和退化问题;另外,提出了一种多分支权重算法,来改善多分支残差模块中各分支的权重分配问题;并将六个以多分支残差模块为主的结构线性连接,组成HCCR识别网络。该模型在CASIA-HWDB1.0、CASIA-HWDB1.1、ICDAR2013数据集上的识别准确率分别达到了97.77%、97.30%、97.64%,表现出高精度的识别效果。展开更多
脑力负荷识别对提高人机交互操作人员的工作绩效具有重要意义。目前已有研究表明,通过提取脑电(electroencephalogram,EEG)信号的能量特征进行脑力负荷识别取得了不错的分类效果。但该方法只关注到信号的幅度信息,而忽略了相位信息;只...脑力负荷识别对提高人机交互操作人员的工作绩效具有重要意义。目前已有研究表明,通过提取脑电(electroencephalogram,EEG)信号的能量特征进行脑力负荷识别取得了不错的分类效果。但该方法只关注到信号的幅度信息,而忽略了相位信息;只分析了各个通道的频域特征,没有考虑不同通道信号之间的同步关系。为充分考虑不同脑区间的功能连接性,提出一种基于加权相位滞后指数(weighted phase lag index,WPLI)热力图的脑力负荷分类方法。对预处理后的脑电信号计算两两通道间的WPLI并绘制热力图,用于评估不同通道信号之间的相位耦合情况,由此反映不同脑区间的功能连接性。由WPLI热力图可以直观地观察到:在高、低负荷状态下,大脑功能连接性的分布存在明显差异。通过实验分别对能量特征图和WPLI热力图采用方向梯度直方图-支持向量机(histogram of oriented gradient-support vector machine,HOG-SVM)和LeNet-5两种方法进行分类。结果表明:WPLI热力图和LeNet-5的组合具有较好的分类结果。展开更多
Given a doubling weightωon the unit disk D,let A_(ω)^(p) be the space of all the holomorphic functions f,where∥f∥A_(ω)^(p):=(∫_(D)|f(z)|_(p)ω(z)dA(z))^(1/p)<∞.We completely characterize the topological conn...Given a doubling weightωon the unit disk D,let A_(ω)^(p) be the space of all the holomorphic functions f,where∥f∥A_(ω)^(p):=(∫_(D)|f(z)|_(p)ω(z)dA(z))^(1/p)<∞.We completely characterize the topological connectedness of the set of composition operators on A_(ω)^(p).As an application,we construct an interesting example which reveals that two composition operators on A_(α)^(p) in the same path component may fail to have a compact difference and give a negative answer to the Shapiro-Sundberg question in the(standard)weighted Bergman space.In addition,we completely describe the central compactness of any finite linear combinations of composition operators on A_(ω)^(p) in three terms:a Julia-Carathéodory-type function-theoretic characterization,a power-type characterization,and a Carleson-type measure-theoretic characterization.展开更多
保护集中连片的优质耕地对于保障国家粮食安全具有重要意义。本文以江西省兴国县为例,从耕地的自然质量、立地条件、生态条件3个层面系统构建耕地综合质量评价指标体系;基于TOPSIS(Technique for order preference by similarity to an ...保护集中连片的优质耕地对于保障国家粮食安全具有重要意义。本文以江西省兴国县为例,从耕地的自然质量、立地条件、生态条件3个层面系统构建耕地综合质量评价指标体系;基于TOPSIS(Technique for order preference by similarity to an ideal solution)综合评价法对耕地综合质量进行评价,引入耕地连片度、耕地空间连通格局分析了耕地空间集聚特征,综合耕地质量及其连通特征划定永久基本农田。结果表明:从耕地综合质量来看,兴国县耕地质量由高到低划分为4个质量等级,面积分别为6 204.95、16 031.72、19 321.79、3 573.76 hm^(2),占总耕地面积的13.75%、35.52%、42.81%、7.92%。总体来看,兴国县中等质量耕地居多,占比为78.33%。从耕地连通性来看,兴国县耕地连片程度由高到低划分为5个等级,其中一级~五级连片耕地面积分别为24 731.44、6 199.73、3 131.54、7 397.71、3 671.80 hm^(2),分别占耕地总面积的54.80%、13.73%、6.94%、16.39%、8.14%,耕地存在不同程度破碎化。将耕地质量三等以上、连片程度四级以上耕地划入基本农田,面积为37 029.62 hm^(2),占耕地总面积的82.05%,与原有划定基本农田相比,实现了划定后永久基本农田“总体稳定、布局优化、质量有提升”的目标。展开更多
文摘In this paper, based on new Lyapunov function, the asymptotic properties of the dynamic neural system with asymmetric connection weights are investigated. Since the dynamic neural system with asymmetric connection weights is more general than that with symmetric ones, the new results are significant in both theory and applications. Specially the new result can cover the asymptotic stability results of linear systems as special cases.
文摘This paper derives some sufficient conditions for exponential stability for the equilibrium point by dividing the state variables of the system according to the characters of the neural networks. The new conditions are described by some blocks of the interconnection matrix. An example is given to demonstrate the effectiveness of the proposed theory.
文摘The bending strength of carbon?fiber/thermoplastic epoxy composites?(CF/TP-EP Compo.)?had?bi-linear increasewith increase of weight-average molecular weight (Mw) of matrix. The transition in the bending strength appeared at around 55k of Mw (“k”?means 103). SEM observation of fractured surface of CF/TP-EP Compo. showed that the fracture mode changed from interfacial failure to fiber breakage dominated failure. The smooth surface of carbon fibers appeared at lower Mw than 55k while some resin remained on the fibers indicating good adhesion between carbon fiber and matrix at higher Mw than 55k. The interfacial shear strength between carbon fiber and matrix bi-linearly increased with an increase of Mw similarly to the bending strength of the composite, measured by the micro droplet test. The dynamic loss tanδ?of the matrix measured at 2?Hz also showed a bi-linear relationship with respect to Mw having a knee point at Mw = 55k. The connection probability of two cracks introduced on?each side of specimens also confirmed that the interfacial strength between carbon fiber and matrix is the key for the mechanical performance of CF/TP-EP Compo. in bending.
基金Supported by the National Science Foundation of China under Grant Nos 60904065,71031007 and 71101013the Program for New Century Excellent Talents in University under Grant No NCET-12-0141.
文摘Natural connectivity has been recently proposed to efficiently characterize the structural robustness of complex networks.The natural connectivity,interpreted as the Helmholtz free energy of a network,can be derived from the graph spectrum.We extend the concept of natural connectivity to weighted complex networks,in which the weight represents the number of multiple edges.We prove that the weighted natural connectivity changes monotonically when the weights are increased or decreased.We investigate the influence of weight on the network robustness within scenarios of weight changing and show that the weighted natural connectivity allows a precise quantitative analysis of the structural robustness for weighted complex networks.
文摘针对容迟网络中节点存储空间有限从而出现大量丢包情况的问题,本文提出了一种适用于节点环境状态的拥塞控制管理策略(congestion control management strategy suitable for the state of the node environment,NEMS):该策略由节点间位置差异相关的控制保留策略和节点自差异相关的丢包策略构成.首先根据节点剩余缓存空间的大小判断节点是处于忙碌状态(BS)还是崩溃状态(CS),当节点处于BS状态时,节点间位置差异相关的控制保留策略结合门限度和连接活跃值的概念分析是否留存新消息.节点自差异相关的丢包策略利用熵权法动态计算各个节点中不同消息属性权重,得到消息丢弃优先级.当拥塞发生,优先删除丢弃优先级大的消息.同时,引入ACK反馈机制,消除网络中已成功投递的冗余消息.仿真结果表明,与无缓存管理的路由策略相比,加入NEMS后使得消息成功投递率增加了约100.93%~200.67%,网络负载降低了约76.05%~81.42%.
文摘脱机手写中文字符识别(handwritten Chinese character recognition,HCCR)在计算机视觉领域一直是一个巨大的挑战。相比传统方法,基于深度学习的网络通过训练大量数据在识别任务中取得了差异化的效果,但识别效果依旧处于发展过程中。基于此,结合DW卷积和残差连接设计了一种多分支残差模块,该模块通过DW卷积以较小的内存和参数量为代价来加深网络深度,增强网络的特征提取能力;再通过残差连接抑制网络梯度问题和退化问题;另外,提出了一种多分支权重算法,来改善多分支残差模块中各分支的权重分配问题;并将六个以多分支残差模块为主的结构线性连接,组成HCCR识别网络。该模型在CASIA-HWDB1.0、CASIA-HWDB1.1、ICDAR2013数据集上的识别准确率分别达到了97.77%、97.30%、97.64%,表现出高精度的识别效果。
文摘脑力负荷识别对提高人机交互操作人员的工作绩效具有重要意义。目前已有研究表明,通过提取脑电(electroencephalogram,EEG)信号的能量特征进行脑力负荷识别取得了不错的分类效果。但该方法只关注到信号的幅度信息,而忽略了相位信息;只分析了各个通道的频域特征,没有考虑不同通道信号之间的同步关系。为充分考虑不同脑区间的功能连接性,提出一种基于加权相位滞后指数(weighted phase lag index,WPLI)热力图的脑力负荷分类方法。对预处理后的脑电信号计算两两通道间的WPLI并绘制热力图,用于评估不同通道信号之间的相位耦合情况,由此反映不同脑区间的功能连接性。由WPLI热力图可以直观地观察到:在高、低负荷状态下,大脑功能连接性的分布存在明显差异。通过实验分别对能量特征图和WPLI热力图采用方向梯度直方图-支持向量机(histogram of oriented gradient-support vector machine,HOG-SVM)和LeNet-5两种方法进行分类。结果表明:WPLI热力图和LeNet-5的组合具有较好的分类结果。
基金supported by National Natural Science Foundation of China (Grant Nos. 12101467 and 12171373)。
文摘Given a doubling weightωon the unit disk D,let A_(ω)^(p) be the space of all the holomorphic functions f,where∥f∥A_(ω)^(p):=(∫_(D)|f(z)|_(p)ω(z)dA(z))^(1/p)<∞.We completely characterize the topological connectedness of the set of composition operators on A_(ω)^(p).As an application,we construct an interesting example which reveals that two composition operators on A_(α)^(p) in the same path component may fail to have a compact difference and give a negative answer to the Shapiro-Sundberg question in the(standard)weighted Bergman space.In addition,we completely describe the central compactness of any finite linear combinations of composition operators on A_(ω)^(p) in three terms:a Julia-Carathéodory-type function-theoretic characterization,a power-type characterization,and a Carleson-type measure-theoretic characterization.
文摘保护集中连片的优质耕地对于保障国家粮食安全具有重要意义。本文以江西省兴国县为例,从耕地的自然质量、立地条件、生态条件3个层面系统构建耕地综合质量评价指标体系;基于TOPSIS(Technique for order preference by similarity to an ideal solution)综合评价法对耕地综合质量进行评价,引入耕地连片度、耕地空间连通格局分析了耕地空间集聚特征,综合耕地质量及其连通特征划定永久基本农田。结果表明:从耕地综合质量来看,兴国县耕地质量由高到低划分为4个质量等级,面积分别为6 204.95、16 031.72、19 321.79、3 573.76 hm^(2),占总耕地面积的13.75%、35.52%、42.81%、7.92%。总体来看,兴国县中等质量耕地居多,占比为78.33%。从耕地连通性来看,兴国县耕地连片程度由高到低划分为5个等级,其中一级~五级连片耕地面积分别为24 731.44、6 199.73、3 131.54、7 397.71、3 671.80 hm^(2),分别占耕地总面积的54.80%、13.73%、6.94%、16.39%、8.14%,耕地存在不同程度破碎化。将耕地质量三等以上、连片程度四级以上耕地划入基本农田,面积为37 029.62 hm^(2),占耕地总面积的82.05%,与原有划定基本农田相比,实现了划定后永久基本农田“总体稳定、布局优化、质量有提升”的目标。
