多目标线性规划的神经网络方法

本文提出了求解多目标线尾规划问题的神经网络，证明了该网络全局稳定于平衡点，且该平衡就是多目标线性规划问题的非劣解，另外它还具有模型小的优点。     

凸规化法实现多维稀疏阵的方向图综合

本文研究了利用凸规划方法综合稀疏随机阵方向图的问题,采用最大信号干扰加噪声比准则,利用附加的线性约束对旁瓣作进一步控制,将方向图设计问题分别转化为二次规划和二次锥规划形式,再用内点法求解。该方法应用于不同形式的稀疏阵列方向图综合,获得了良好的结果。     

线性约束条件下任意凸函数的神经网络优化模型

该文提出了线性约束条件下任意凸函数的神经网络优化模型,所构造的能量函数的平衡点即为原问题的最优解,克服了传统的神经网络优化方法所存在的问题,网络是全局稳定的,并能收敛到最优点,计算机仿真结果证明了本文方法的有效性。     

两跳放大转发中继网络中的ε-全局最优多中继选择策略

研究了并行两跳放大转发中继网络中的多中继选择策略问题。以接收端的信噪比(signal-to-noiseratio,SNR)最大化为目标,该问题首先被规划为0-1非线性整数规划问题,这是一个NP-hard问题,其最优解只可以通过穷举方法得到,但是对于节点数量很大的网络,低效的穷举方法是无法容忍的。基于分支定界(branch-and-bound)框架并结合线性变形放松技术(reformulation-linearization technique,RLT),提出了一种ε-全局最优的多中继选择策略。仿真结果表明该方法能很好地以预定精度ε逼近全局最优解。通过此方法得到的性能可以作为其他算法的参考基准。     

解框形约束最小二乘问题的神经网络

将Ｈｏｐｆｉｅｌｄ神经网络用于解最优化问题，给出了一个解框形约束小二乘问题的离散形神经网络，当这一网络达到其稳定状态，即其能量函数为最小时，由网络的输出可以获得问题的最优解。     

二次型规划问题的进一步研究

进一步分析了求解二次型规划问题的神经网络方法，给出了这一方法的稳定性的可检验条件，并证明在这个条件下网络是全局收敛的。最后，给出了数值模拟例子。     

一种三维刚体行动参数鲁棒估计的神经算法

该文提出一种在于特征点匹配的刚体运动参数估计方法,在运动估计线性算法的基础上,文中利用全最小二乘（ＴＬＳ）方法来进行求解,并建立次分量提取神经元来获得该全最小二乘解,基于测量数据中出格点（Ｏｕｔｌｉｅｒ）的存在,我们在神经元的权值学习规划中引入鲁棒估计思想。实验结果表明,该方法能有效地克服同格点产生的误差,准确在估计出刚体的三维运动参数,比较令人满意。     

求解一类特殊优化问题的模拟人工神经网络方法

在许多实际应用中，经常遇到一类特殊的非线性规划问题，即具有特殊二次约束的二次规划问题，本文提出了一种求解该ＱＰＱＣ问题的模拟反馈人工神经网方法，详细分析了这种网络方法的性能。最后给出我们的一些计算结果。     

解决二次规划的一中立型投影神经网络

在这篇文章中我们构造了解决线性交分不等式和二次规划问题的一类中立型投影神经网络,在这篇文章里我们主要用的是微分方程和放缩的方法证明我们构造的投影神经网络全局指数稳定的.而且结果分两种情况来讨论,两种情况下系统都是指数稳定的,因此它可以用来解决一系列约束最优化问题的二次规划问题和线性变分不等式等问题.     

以"毒"攻毒型蠕虫病毒的研究

本文在对一支典型攻毒型蠕虫进行性能分析的基础上,对其不足之处进行改进,提出 了一种新的概念型攻毒蠕虫的概念及技术。该新蠕虫力图找到网络性能影响最小和传播速度最大的 平衡点,以便达到在网络性能降低最小的情况下获得网络安全整体性提高的目的。     

Channel assignment for cellular radio using extended dynamic programming

The channel assignment is an important aspect of cellular radio networks. Because of the limitations on the frequency spectrum, the optimal or near-optimal channel assignment has become an essential part of the network operations of wireless personal communication systems. We formulate a new strategy for the channel assignment problem in agreement with the electromagnetic compatibility constraints. We introduce and formulate the extended dynamic programming (EDP), as an extension of dynamic programming for solving the channel assignment problem in a cellular system. Using EDP an algorithm is developed for fixed channel assignment problem and it is tested and compared with other existing methods by solving different problems. In agreement with electromagnetic compatibility constraints, solution strategy based on EDP algorithm finds many valid solutions with minimum possible bandwidth.     

多目标动态规划的神经网络方法

本文提出了求解多目标动态规划的神经网络方法,证明了该网络是Ｌｙａｐｕｎｏｖ稳定的,且在适当的条件下收敛于多目标动态规划的（局部）非劣解,最后,文中给出了算例及其模拟结果。     

一种求解带等式约束的二次规划问题的新神经网络模型

本文提出了一种新的神经网络模型用以求解带等式约束的二次规划问题。本模型大大改进了Shenguei zhang等人(1992)给出的二阶神经网络。后者的结构复杂,且只适用于求解具有唯一最优解的情况,而不能求解最优解为无界集合的情形。与它相比,本文提出的新神经网络模型不仅能够精确求解具有唯一最优解的情况,而且还能实时求解最优解集合为无界的情况,此外,在结构上也大为简化,易于电路实现。     

On Neural Network Approach to Computing the Quadratic Programming Problem：a Further Study

ＯｎＮｅｕｒａｌＮｅｔｗｏｒｋＡｐｐｒｏａｃｈｔｏＣｏｍｐｕｔｉｎｇｔｈｅＱｕａｄｒａｔｉｃＰｒｏｇｒａｍｍｉｎｇＰｒｏｂｌｅｍ：ａＦｕｒｔｈｅｒＳｔｕｄｙＸｉａＹｏｕｓｈｅｎａｎｄＷｕＸｉｎｙｕ（ＮａｎｊｉｎｇＵｎｉｖｅｒｓｉｔｙｏｆＰｏｓｔｓａｎ...     

用动态规划方法求解最短运输路线问题

在日常生活中,很多领域都要涉及到最短路径问题,如何求解最短路径的问题是非常重要的。阐述了动态规划方法的基本思想、求解最短路径问题的步骤,并使用C语言编程解决了最短运输路线问题,说明了动态规划求解最短路径的可行性。     

Joint routing and dimensioning of optical burst switching networks

Existing methods for handling routing and dimensioning in dynamic WDM networks solve the two problems separately. The main drawback of this approach is that a global minimum cost solution cannot be guaranteed. Given that wavelengths are costly resources, determining the minimum network cost is of fundamental importance. We propose an approach which jointly solves the routing and dimensioning problems in optical burst switching (OBS) networks, guaranteeing a target blocking per connection. The method finds the set of routes and the number of wavelengths per network link that minimise the total network cost. To accomplish this, an integer linear programming problem is solved. The proposed method was applied to ring networks, where the optimal solution achieves a reduction in the network cost of 10–40% (for traffic loads <0.4, compared to solving both problems separately). In the case of mesh topologies, to reduce the computational complexity of the method, we applied a variation of it which achieves a local minimum. Even so, a reduction of 5–20% (for traffic loads <0.4) in the network cost was obtained. This ability to lower network cost could make the proposed method the best choice to date for dynamic network operators.     

一种网络备用资源的集中规划算法

满足多级生存性要求的网络带宽资源规划问题是一个存在多个非线性不等式约束的整数规划问题,本文提出了一种考虑移动工作通道对胆识物资源规划算法,并采用拉格朗日松驰法和次梯度法求解该规划问题。     

计算机通信网中路由选择和容量分配问题的遗传算法求解

计算机通信网中，对路由选择和容量分配问题进行综合优化设计时的数学模型，是一个多约束条件的非线性０－１规划。本文以遗传算法的基本思想为基础，设计了一个求解该问题的遗传寻优算法，对计算机通信网优化设计实例的计算表明，这一方法能够迅速求出问题的全局近似最优解，并具有高的计算精确度。     

Optimal channel assignment in cellular networks

In this paper, we consider the problem of assigning frequencies to mobile terminals in a cellular network. We show that an optimal solution can be obtained by solving a sequence of alternating linear and quadratic maximization programming problems. We address co-channel constraints and adopt as an objective function the maximization of potentially established calls. Our algorithm is fairly general, and does not depend on any special network structure. This study indicates that mathematical programming can be used as an efficient technique for solving the aforementioned problem.     

An Efficient Heuristic for Solving an Extended Capacitated Concentrator Location Problem

In this paper, a mathematical model and a solution algorithm are developed for solving an extended capacitated concentrator location problem. Our model extends the conventional formulation by simultaneously addressing the two capacity constraints, total connection ports and maximum data processing rate, on each concentrator to be selected for satisfying the communication demands of the given end-user nodes. Since the problem is NP-complete, an efficient and effective Lagrangian heuristic is developed and tested by solving 100 randomly generated test problems with sizes ranging from 30(nodes)×30(concentrators) to150×30. Altogether 58% of the tested problems are solved optimally with an average solution gap 0.36% from the optimality and average solution times are from a few seconds to one half of a minute.