Segment-based adaptive hyper-Erlang model for long-tailed network traffic approximation

Modeling the long-tailedness property of network traffic with phase-type distributions is a powerful means to facilitate the consequent performance evaluation and queuing based analysis. This paper improves the recently proposed Fixed Hyper-Erlang model (FHE) by introducing an adaptive framework (Adaptive Hyper-Erlang model, AHE) to determine the crucially performance-sensitive model parameters. The adaptive model fits long-tailed traffic data set directly with a mixed Erlang distribution in a new divide-and-conquer manner. Compared with the well-known hyperexponential based models and the Fixed Hyper-Erlang model, the Adaptive Hyper-Erlang model is more flexible and practicable in addition to its accuracy in fitting the tail behavior.

Continuous State Dynamic Programming via Nonexpansive Approximation

This paper studies fitted value iteration for continuous state numerical dynamic programming using nonexpansive function approximators. A number of approximation schemes are discussed. The main contribution is to provide error bounds for approximate optimal policies generated by the value iteration algorithm.      

Lower estimation of approximation rate for neural networks

Let SFd and Πψ,n,d = { nj=1bjψ(ωj·x+θj) :bj,θj∈R,ωj∈Rd} be the set of periodic and Lebesgue’s square-integrable functions and the set of feedforward neural network (FNN) functions, respectively. Denote by dist (SF d, Πψ,n,d) the deviation of the set SF d from the set Πψ,n,d. A main purpose of this paper is to estimate the deviation. In particular, based on the Fourier transforms and the theory of approximation, a lower estimation for dist (SFd, Πψ,n,d) is proved. That is, dist(SF d, Πψ,n,d) (nlogC2n)1/2 . T...     

Sampled-data based average consensus with measurement noises: convergence analysis and uncertainty principle

In this paper, sampled-data based average-consensus control is considered for networks consisting of continuous-time first-order integrator agents in a noisy distributed communication environment. The impact of the sampling size and the number of network nodes on the system performances is analyzed. The control input of each agent can only use information measured at the sampling instants from its neighborhood rather than the complete continuous process, and the measurements of its neighbors’ states are corrupted by random noises. By probability limit theory and the property of graph Laplacian matrix, it is shown that for a connected network, the static mean square error between the individual state and the average of the initial states of all agents can be made arbitrarily small, provided the sampling size is sufficiently small. Furthermore, by properly choosing the consensus gains, almost sure consensus can be achieved. It is worth pointing out that an uncertainty principle of Gaussian networks is obtained, which implies that in the case of white Gaussian noises, no matter what the sampling size is, the product of the steady-state and transient performance indices is always equal to or larger than a constant depending on the noise intensity, network topology and the number of network nodes.     

基于相似关系的决策系统的知识获取算法

本文基于一种相似关系下的粗糙集模型,给出了一些基本性质,讨论了该相似关系下的决策系统的知识获取算法。该算法可以避免因数据离散化的处理而产生的信息系统相似度的误差。     

最大简单共享问题的快速近似算法

介绍了一种基于复制结点的消除线路交叉的模型.该模型提出了一个优化问题,就是最小化结点复制的数量.同时提出一个自定义问题——"最大简单共享问题",并证明最小化结点复制的数量与最大共享问题是等价的.证明了最大简单共享问题是NP-hard的,给出了一种简单的贪心算法,并证明该贪心算法的近似度为3.引入一个"最大互斥简单共享问题",该问题是最大简单共享问题的2-近似.将其转化为在一系列图上的完美匹配问题,使该问题可以在多项式时间内得到完美解决.最后,在最大互斥简单共享的基础上,用局部搜索的方法将近似度提高到12/7.     

方位向时变非均匀采样的合成孔径声纳成像

采用多接收阵技术能够使合成孔径声纳的测绘速率和测绘带同时得到提高,但带来了方位向非均匀采样以及不满足停走停假设的问题.本文建立了高测绘速率、宽测绘带条件下不做停走停近似的多接收阵合成孔径声纳信号模型,采用频域方法对合成孔径声纳时变的非均匀方位向信号进行了重构,并导出了重构方法的快速算法.重构方法解决了高测绘速率、宽测绘带条件下的偏移及散焦问题.仿真和实验验证了该方法的可行性.     

T-S模糊系统的构造及其逼近能力的充分条件

针对任意多变量非线性函数, 至今缺少一种简单有效的构造性方法来建造一个T-S模糊系统对其实现给定精度的模糊逼近. 本文在逼近函数g(X)∈C2及g(X)不∈C2两种情况下给出了一种当模糊集满足均匀性、连续性、正规性条件下T-S模糊系统的构造方法以及该T-S模糊系统逼近能力的充分条件. 该构造性方法简单易行且其逼近能力充分条件的保守性低. 最后通过仿真研究证实了该方法的正确性与有效性.     

用改进的竞争Hopfield神经网络求解多边形近似问题

多边形近似是提取曲线特征点和简化曲线描述的一种重要方法.提出一种改进的Hopfield神经网络多边形近似算法,该算法利用选择拐点策略减少了搜索空间,重新定义了神经网络的能量函数,使其更能反映优化目标;引?入合并拆分搜索策略,有效帮助神经网络脱离局部最小值.实验结果表明,提出的改进算法是有效的,比其它算法如关键点检测法、竞争Hopfield神经网络、混沌Hopfield神经网络、遗传算法等具有更优的性能.     

An augmented Lagrangian approach to non-convex SAO using diagonal quadratic approximations

Successful gradient-based sequential approximate optimization (SAO) algorithms in simulation-based optimization typically use convex separable approximations. Convex approximations may however not be very efficient if the true objective function and/or the constraints are concave. Using diagonal quadratic approximations, we show that non-convex approximations may indeed require significantly fewer iterations than their convex counterparts. The nonconvex subproblems are solved using an augmented Lagrangian (AL) strategy, rather than the Falk-dual, which is the norm in SAO based on convex subproblems. The results suggest that transformation of large-scale optimization problems with only a few constraints to a dual form via convexification need sometimes not be required, since this may equally well be done using an AL formulation.