直接网络中的自适应路由算法分析

互连网络是大规模并行计算机的重要组成部分，路由算法是其中决定网络性能的重要因素，本文在直接网络结构基础上对路由算法进行讨论，给出了一种分类方法，并着重对采用虫孔路由开关技术的自适应路由算法进行分析，为进一步的评价和设计新的算法提供了参考。     

基于神经网络的非线性控制系统自组织辨识

本文首先将自组织神经网络算法向一般化情形引伸，接着把自组织过程应用到一般非线性系统的动态过程分类，使得整个非线性系统能够按照输入输出样本空间的概率密度自组织，成为许多具有不同分类核心和感受野的线性子空间逼近。在此基础上，我们采用通用最小二乘算法，以子空间的非线性问题线性化误差作为依据，并进一步运用自组织神经网络的合作与竞争思想，最终得到一般情形的非线性系统的最小二乘辨识。仿真结果表明了本方法的可行性与优越性。     

神经网络的函数逼近理论

分析了将函数逼近理论与方法引入神经网络研究的必要性；从经典函数逼近与统计分析两方面详细地讨论了多层前馈网（ＭＬＰ）逼近能力分析的基本方法及结论；分析了正则理论观点下的径向基函数网络（ＲＢＦ）的逼近能力；讨论了ＲＢＦ网与多层前馈网在最佳逼近特性上的差异。文末指出了神经网络函数逼近的发展方向。     

支持QoS的无线Mesh网络机会路由优化算法

针对空中骨干Mesh网络资源有限、计算能力相对不足的特点以及传统简单机会路由(Simple Opportunistic Adaptive Routing,SOAR)路由算法未充分考虑负载均衡与不同业务服务质量(Quality of Service,Qo S)保障需求差异性的问题,提出一种支持业务区分的改进型SOAR路由算法。该算法在考虑链路拥塞控制和负载均衡的基础上,定义综合预期传输次数来描述链路的综合状态,有效降低网络拥塞概率;同时根据传输业务类型的不同,设计一种基于层次分析法的路由选择策略,实现路径选择与业务类型的动态匹配。仿真结果表明,在重负载条件下,改进型SOAR路由算法相比传统SOAR路由算法其时延、吞吐量和吞吐率性能明显提升。当网络中存在不同类型业务时,改进型SOAR路由算法能够根据业务Qo S保障需求的差异性自适应选择最佳传输路径。     

非结构动网格分区并行计算方法

针对包含运动边界的非定常流动问题,基于弹簧近似法设计了一套非结构动网格分区并行计算流程,引入了METIS软件进行非结构网格二次剖分,研究了对接边界通信类型及其处理方法,基于消息传递并行编程模式实现了非结构动网格求解器的并行化。测试结果表明:该求解器具有较高的并行效率,网格变形模块对非结构动网格并行计算性能具有重要影响。     

扩展卡尔曼滤波的有向传感器网络移动目标跟踪算法

针对传感器网络中跟踪目标需要大量的节点协同工作,还需要实时处理和传输大量数据,提出一种基于扩展卡尔曼滤波的有向传感器网络目标跟踪算法(EK-MTDC),重点研究了传感器网络中的扇区数量对节点间数据传输与目标跟踪精度的影响,根据对目标状态的分析,通过压缩参与监测的节点个数,选择激活网络中节点相交区域内的节点对跟踪目标进行监测。仿真结果表明,该算法能在不降低跟踪效果的前提下,降低网络能耗,延长其使用寿命。     

Analysis of production lines with Coxian service times and no intermediate buffers

This paper uses the holding time model (HTM) method to derive an approximate analytic formula for the calculation of the mean throughput of a K-station production line with no buffers between any two successive stations. Service times follow the two-stage Coxian (C²) distribution at all stations. The paper provides a formula that relates the third moment of the service completion (or virtual service) time with the respective parameters of the service time, the repair time and the time to breakdown (the latter is assumed to follow the exponential distribution). In this way, it concludes that under certain conditions the two-stage Coxian distribution can be used to approximate any general distribution matching the first three moments of the service completion time distribution. The mean holding times (consisting of the service and blocking periods) of all stations of the line are obtained in an analytical form. Numerical results are provided for the mean throughput of lines with up to 20 stations. These results are shown to have a good accuracy compared against results obtained from the Markovian state method (for short lines) and results from simulation (for longer lines). © 1998 John Wiley & Sons, Inc. Naval Research Logistics 45: 669–685, 1998     

The effect of training data set size and the complexity of the separation function on neural network classification capability: The two-group case

Classification among groups is a crucial problem in managerial decision making. Classification techniques are used in: identifying stressed firms, classifying among consumer types, and rating of firms' bonds, etc. Neural networks are recognized as important and emerging methodologies in the area of classification. In this paper, we study the effect of training sample size and the neural network topology on the classification capability of neural networks. We also compare neural network capabilities with those of commonly used statistical methodologies. Experiments were designed and carried out on two-group classification problems to find answers to these questions. The prediction capability of the neural network models are better than traditional statistical models. The learning capability of the neural networks is improving compared to traditional models because the discriminate function is more complex. For real world classification problems, the usage of neural networks is highly recommended, for two reasons: learning capability and flexibility. Learning capability: Neural network classifies better in sterile experiments as performed in this research. Flexibility: Real life data are rarely not contaminated with noise, such as unknown distributions, and missing variables, etc. Neural networks differ from a statistical model that it is not dependent on any assumption concerning the data set distribution. © 1997 John Wiley & Sons, Inc. Naval Research Logistics 44: 699–717, 1997     

基于RBF网络的导弹攻击区解算及对比分析

研究了神经网络在空空导弹攻击区处理中的运用,在提出多层前馈(MF)网络和径向基函数(RBF)网络设计方案的基础上,着重对MF网络的BP算法、改进BP算法以及RBF网络的应用进行了仿真,并从收敛速度、逼近精度和实时性三个方面,对计算结果进行了对比分析,结论表明RBF网络在导弹攻击区处理中具有更高的可行性,更适用于快速高精度的火控解算。     

一种基于神经网络的磁性目标定位方法

磁性目标定位问题可归结为一类非线性规划问题的求解 .该问题的最大特点就是其目标函数的计算过程极为繁琐 .而目标函数计算的快慢对磁定位的实时性有很大影响 .在详细研究了目标函数之后 ,给出了目标函数的一种神经网络结构实现 .由于该网络结构易于并行计算和VLSI实现 ,从而可使磁定位的实时性得到改善 .