Performance & packet traffic dynamics of Packet Switching Network model

Dynamics of packet traffic in data communication networks can be complex and often not well understood. Understanding of these complex dynamics is important for their control, prediction purposes and for the data networks design. The engineering community has described wired data networks architectures and studied them by means of a layered, hierarchical abstraction called ISO OSI (International Standard Organization Open System Interconnect) Reference Model. The Network Layer of the ISO OSI Reference Model is responsible for routing packets across the network from their sources to their destinations and for control of congestion in data networks. Using an abstraction of the Network Layer that we developed, we investigate packet traffic dynamics in our data network models of data communication networks of packet switching type, in particular near the phase transition point from free flow to congestion. We explore how these dynamics and network performance indicators are affected by network connection topology and routing algorithms. We consider static and adaptive routing algorithms. (© 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim)     

偏最小二乘回归的贝叶斯正则化神经网络集成模型在证券分析预测中的应用

神经网络集成技术能有效地提高神经网络的预测精度和泛化能力,已经成为机器学习和神经计算领域的一个研究热点.利用Bagging技术和不同的神经网络算法生成集成个体,并用偏最小二乘回归方法从中提取集成因子,再利用贝叶斯正则化神经网络对其集成,以此建立上证指数预测模型.通过上证指数开、收盘价进行实例分析,计算结果表明该方法预测精度高、稳定性好.     

Modelling of dynamic networks of thin elastic plates

The purpose of this paper is to derive junction conditions for networks of thin elastic plates and to analyse the dynamic equations of such networks. Junction conditions for networks of Kirchhoff plates and networks of Reissner–Mindlin plates are derived based on geometric considerations of the deformation at a junction. It is proved that the dynamic system which describes the Reissner–Mindlin network is well-posed is an appropriate energy space. It is further established that the Kirchhoff network is obtained in the limit of the Reissner–Mindlin network as the shear moduli go to infinity.     

Prediction of swerving motion of a dual-spin projectile with lateral pulse jets in atmospheric flight

Using the linear theory for a dual-spin projectile in atmospheric flight, closed form expressions are obtained for swerving motion under the action of lateral pulse jets. Trajectory results generated by the linear theory equations and a fully nonlinear seven degree-of-freedom dual spin projectile model agree favorably. The analytic solution provides a relatively straightforward and computationally efficient means of trajectory estimation which is useful within smart weapon flight control systems. In order to accurately predict the impact point using the analytic solution, the dual-spin projectile linear model must be updated periodically. Terminal impact point prediction degrades rapidly as the linear model update interval is increased beyond a critical value. Control authority, as defined by the change in impact location due to a pulse jet firing, steadily decreases as a function of projectile down range position.     

A Learning Framework for Neural Networks Using Constrained Optimization Methods

Conventional supervised learning in neural networks is carried out by performing unconstrained minimization of a suitably defined cost function. This approach has certain drawbacks, which can be overcome by incorporating additional knowledge in the training formalism. In this paper, two types of such additional knowledge are examined: Network specific knowledge (associated with the neural network irrespectively of the problem whose solution is sought) or problem specific knowledge (which helps to solve a specific learning task). A constrained optimization framework is introduced for incorporating these types of knowledge into the learning formalism. We present three examples of improvement in the learning behaviour of neural networks using additional knowledge in the context of our constrained optimization framework. The two network specific examples are designed to improve convergence and learning speed in the broad class of feedforward networks, while the third problem specific example is related to the efficient factorization of 2-D polynomials using suitably constructed sigma-pi networks.     

基于邻域粗糙集的网络入侵分类诊断组合模型研究

网络入侵诊断直接影响网络正常运行和安全.针对入侵类型复杂,现有分类诊断模型精度有限的问题,提出一种基于邻域粗糙集的网络入侵分类诊断优化模型.首先,运用邻域粗糙集对网络入侵数据进行条件属性的约简,确定关键属性,然后将其作为训练输入构建相关向量机分类诊断模型,并同时运用遗传算法进行超参数优化,提高模型诊断精度和速度.通过KDDCup99数据集对优化模型性能进行检验,结果表明,组合预测方法精确度高于支持向量机、相关向量机和BP神经网络.组合模型诊断精度高、速度快,具有优异的综合性能.     

Percolation of interdependent network of networks

Complex networks appear in almost every aspect of science and technology. Previous work in network theory has focused primarily on analyzing single networks that do not interact with other networks, despite the fact that many real-world networks interact with and depend on each other. Very recently an analytical framework for studying the percolation properties of interacting networks has been introduced. Here we review the analytical framework and the results for percolation laws for a Network Of Networks (NONs) formed by n interdependent random networks. The percolation properties of a network of networks differ greatly from those of single isolated networks. In particular, because the constituent networks of a NON are connected by node dependencies, a NON is subject to cascading failure. When there is strong interdependent coupling between networks, the percolation transition is discontinuous (first-order) phase transition, unlike the well-known continuous second-order transition in single isolated networks. Moreover, although networks with broader degree distributions, e.g., scale-free networks, are more robust when analyzed as single networks, they become more vulnerable in a NON. We also review the effect of space embedding on network vulnerability. It is shown that for spatially embedded networks any finite fraction of dependency nodes will lead to abrupt transition.     

基于复杂网络演化博弈的企业低碳创新合作行为网络演化机理研究

针对企业低碳创新合作所面临的复杂问题,基于现实复杂网络结构特征,运用演化博弈理论研究有限理性下企业低碳创新合作行为网络演化机理,利用Matlab仿真技术探究无标度网络载体上微观因素对低碳创新合作行为的影响。研究结果表明:低碳创新利益分配、协同效益和违约惩罚对低碳创新合作行为网络演化结果的影响最为显著,网络规模越大网络演化速度越慢,网络规模越小对协同系数和利益分配系数的敏感性越强,网络规模越大对技术溢出系数和违约惩罚的敏感性越强。研究结论可以为企业低碳创新合作策略制定提供解决依据。     

资源有限网络计划启发式方法的评价（上）：理论分析与试验设计

在对网络计划问题分析的基础上,提出了两个衡量资源约束网络计划问题复杂性特征的度量指标-网络复杂性系数CNC及有资源约束网络的复杂性系数CRNC,并基于此设计了一种产生给定网络复杂性要求的随机活动网络发生器GRAN,最后针对资源约束的网络计划问题设计了一种评价与分析其启发式算法效果的试验模型。     

Initial states iterative learning for three-dimensional ballistic endpoint control

In this paper, an initial states iterative learning control algorithm is proposed for control of the ballistic endpoint displacement in three-dimensional space, where the target is moving and the projectile experiences system uncertainties. The characteristics of the three-dimensional ballistic process are formulated and explored, and the learning algorithm is proposed in the spatial domain. The algorithm consists of two parts. First, the initial speed and angles are iteratively learned to make the projectile attain a fixed position. Second, the shooting time is learned to tune the arrival time of the projectile. Since the dimensions of the solution space are larger than that of the task space, three control manners, including shooting speed, shooting angle and their combination, are researched respectively. Through rigorously analyzed, it is proved that the algorithm is convergent and the multiple initial states can be adjusted simultaneously. Finally, an example of practical cannonball projection is presented to verify the effectiveness of the proposed algorithms.     

引入技术指标的BP网络在股市预测中的应用

本文使用股市分析中常用的一些技术指标构造BP网络的输入样本向量,在此基础上,对沪市股指的涨跌进行了预测.数值实验结果表明,该方法能够提高网络预测的正确率.使用跳跃学习及三种变学习率、批方式的学习算法对BP网络进行了训练,节省了预测时间.运用“在线预测”的方法对预测过程进行了跟踪.针对预测样本在预测性能及预测结果方面存在的差异,引入预测样本中心距离比的概念对其进行简单的划分,得到一些富有启发性的结果.     

Exploring express delivery networks in China based on complex network theory

Based on the complex network theory, we explore an express delivery system in China, which consists of two delivery networks, namely, the air delivery network (ADN) and the ground delivery network (GDN). Systematic structural analysis indicates that both delivery networks exhibit small‐world phenomenon, disassortative mixing behavior, and rich‐club phenomenon. However, there are significant differences between ADN and GDN in terms of degree distribution property and community structure. On the basis of the Barabási‐Albert model, we have proposed a network model incorporating the structural features of the two delivery networks to reveal their evolutionary mechanisms. Lastly, the parcel strength and the distance strength are analyzed, which, respectively, reflect the number of parcels and the long‐haul delivery distance handled by a node city. The strengths are highly heterogeneous in both delivery networks and have intense correlations with topological structures. These works are beneficial for express enterprises to construct or extend their express delivery networks, and provide some useful insights on improving parcel delivery service. © 2014 Wiley Periodicals, Inc. Complexity 21: 166–179, 2015     

BP-GA混合优化策略在人力资源战略规划中的应用

采用混合优化策略训练神经网络,进而实现地区人力资源数据的时间序列预测.神经网络,尤其是应用反向传播(back propagation,简称BP)算法训练的神经网络,被广泛应用于预测中.但是BP神经网络训练速度慢、容易陷入局部极值.遗传算法(genetic algorithm,简称GA)具有很好的全局寻优性.因而提出将BP和GA结合起来的混合优化策略训练神经网络,来实现人力资源数据预测.与BP算法相比,数值计算结果表明预测精度高、速度快,为地区人力资源数据的时间序列预测研究提供了一条新的途径.     

Bargaining in endogenous trading networks

We study a model in which heterogeneous agents first form a trading network where linking costs are positive but infinitesimally small. Then, a seller and a buyer are randomly selected among the agents to bargain through a chain of intermediaries. We determine both the trading path and the allocation of the surplus among the seller, the buyer and the intermediaries at equilibrium. We show that, under the initiator bargaining protocol, a trading network is pairwise stable if it is a core–periphery network where the core consists of all impatient agents who are linked to each other and the periphery consists of all patient agents who have a single link towards an impatient agent. Once agents do not know the impatience of other agents, each bilateral bargaining session may involve delay. Then, core–periphery networks may not be pairwise stable because agents may prefer to add links for reducing the length of trading paths and so avoiding costly delays in reaching a global agreement.     

Linear unlearning for cross-validation

The leave-one-out cross-validation scheme for generalization assessment of neural network models is computationally expensive due to replicated training sessions. In this paper we suggest linear unlearning of examples as an approach to approximative cross-validation. Further, we discuss the possibility of exploiting the ensemble of networks offered by leave-one-out for performing ensemble predictions. We show that the generalization performance of the equally weighted ensemble predictor is identical to that of the network trained on the whole training set.Numerical experiments on the sunspot time series prediction benchmark demonstrate the potential of the linear unlearning technique.     

Optimization of neural network for ionic conductivity of nanocomposite solid polymer electrolyte system (PEO-LiPF6-EC-CNT)

In this study, the ionic conductivity of a nanocomposite polymer electrolyte system (PEO-LiPF₆-EC-CNT), which has been produced using solution cast technique, is obtained using artificial neural networks approach. Several results have been recorded from experiments in preparation for the training and testing of the network. In the experiments, polyethylene oxide (PEO), lithium hexafluorophosphate (LiPF₆), ethylene carbonate (EC) and carbon nanotubes (CNT) are mixed at various ratios to obtain the highest ionic conductivity. The effects of chemical composition and temperature on the ionic conductivity of the polymer electrolyte system are investigated. Electrical tests reveal that the ionic conductivity of the polymer electrolyte system varies with different chemical compositions and temperatures. In neural networks training, different chemical compositions and temperatures are used as inputs and the ionic conductivities of the resultant polymer electrolytes are used as outputs. The experimental data is used to check the system’s accuracy following the training process. The neural network is found to be successful for the prediction of ionic conductivity of nanocomposite polymer electrolyte system.     

Path relinking and GRG for artificial neural networks

Artificial neural networks (ANN) have been widely used for both classification and prediction. This paper is focused on the prediction problem in which an unknown function is approximated. ANNs can be viewed as models of real systems, built by tuning parameters known as weights. In training the net, the problem is to find the weights that optimize its performance (i.e., to minimize the error over the training set). Although the most popular method for training these networks is back propagation, other optimization methods such as tabu search or scatter search have been successfully applied to solve this problem. In this paper we propose a path relinking implementation to solve the neural network training problem. Our method uses GRG, a gradient-based local NLP solver, as an improvement phase, while previous approaches used simpler local optimizers. The experimentation shows that the proposed procedure can compete with the best-known algorithms in terms of solution quality, consuming a reasonable computational effort.     

突发事件对交通网络的影响评估指标和方法

近年来,由于自然灾害,不断地环境恶化以及不可预测的灾害侵袭,突发事件时有发生,突发事件可能对交通网络产生非常严重的影响。本文给出了一些指标来表征突发事件对交通网络的影响程度。这些指标适用于用户最优、系统最优以及效率损失等情形。 我们给出了这些指标的性质并讨论了在用户最优、系统最优以及效率损失等情形下的指标之间的关系。交通网络的数值算例表明了关于突发事件的指标的合理性。突发事件对交通网络的影响程度评估对于道路规划、道路维护以及网络脆弱性分析均有重要意义。     

How neutral networks influence evolvability

Evolutionary algorithms apply the process of variation, reproduction, and selection to look for an individual capable of solving the task at hand. In order to improve the evolvability of a population we propose to copy important characteristics of nature's search space. Desired characteristics for a genotype–phenotype mapping are described and several highly redundant genotype–phenotype mappings are analyzed in the context of a population‐based search. We show that evolvability, defined as the ability of random variations to sometimes produce improvement, is influenced by the existence of neutral networks in genotype space. Redundant mappings allow the population to spread along the network of neutral mutations and the population is quickly able to recover after a change has occurred. The extent of the neutral networks affects the interconnectivity of the search space and thereby affects evolvability. © 2002 Wiley Periodicals, Inc.