动态直觉模糊多属性决策的VIKOR扩展方法

隐变量是观察不到或虚拟的变量,直接利用数据驱动的学习方法难以有效地发现隐变量,因而需要结合概率图结构分析的方法。针对基于结构分析的隐变量发现方法中难以确定隐变量个数和位置的问题,提出一种基于结构分解和因子分析的隐变量发现算法(S-FAHF)。S-FAHF算法利用联合树算法生成具较强依赖关系的变量子集,利用因子分析思想,通过求变量子集的特征值和累积贡献率确定变量子集中隐变量的个数,利用负荷矩阵确定隐变量的位置,最后利用打分函数测试所发现的隐变量的有效性。通过算法比较和实验结果表明,该方法能准确地确定贝叶斯网络中隐变量的个数及位置。     

多变量广义预测自校正控制隐式算法

本文提出了一种多变量广义预测自校正控制隐式算法,它计算量小,能用于Bo奇异,输入、输出维数不等,非最小相位及开环不隐定等一大类MIMO系统,仿真结果表明,该算法鲁棒性强,且稳态无偏。     

基于惩罚函数泛化的神经网络剪枝算法研究

神经网络的隐层数和隐层节点数决定了网络规模,并对网络性能造成较大影响。在满足网络所需最少隐层节点数的前提下,利用剪枝算法删除某些冗余节点,减少隐层节点数,得到更加精简的网络结构。基于惩罚函数的剪枝算法是在目标函数后加入一个惩罚函数项,该惩罚函数项是一个变量为网络权值的函数。由于惩罚函数中的网络权值变量可以附加一个可调参数,将单一惩罚函数项泛化为一类随参数规律变化的新的惩罚函数,初始惩罚函数可看作泛化后惩罚函数的参数取定值的特殊情况。实验利用基于标准BP神经网络的XOR数据进行测试,得到隐层节点剪枝效果和网络权值随惩罚函数的泛化而发生变化,并从数据分析中得出具有更好剪枝效果及更优网络结构的惩罚函数泛化参数。     

局部因果关系分析的隐变量发现算法

结构分析的隐变量发现方法难以有效地发现隐变量且可解释性较差。基于因果关系和局部结构的不确定性,提出了一种基于局部因果关系分析的隐变量发现算法（hidden variable discovering algorithm based on local causality analysis,LCAHD）。LCAHD算法给出了因果结构熵的定义,将因果知识和不确定性知识相融合,以因果关系的不确定性程度作为隐变量存在的判定依据,并对这一依据进行了理论上的论证。LCAHD算法首先通过寻找目标变量的马尔科夫毯来提取局部依赖结构,并基于扰动学习获得扰动数据,联合扰动数据和观测数据学习局部依赖结构中的因果关系;然后利用因果结构熵对局部因果结构中因果关系的不确定性进行度量,并利用隐变量和因果关系不确定性之间的相关性判定条件,确定隐变量的存在性。分别针对标准网络和股票网络进行了实验,结果表明,该算法能准确地确定隐变量的位置,具有较好的解释性。     

BP神经网络结构参数的计算机自动确定

研究表明,由多层FNN的BP算法误差函数构成的非线性方程组的独立方程个数和FNN的待求未知变量的个数应该相等,该方程组才能有唯一组解。由此导出网络结构方程式,进而导出隐层层数判别式和每层神经元个数判别式。依据Kolmogorov定理,由该判别式得出求解FNN隐层层数和每个隐层神经元个数的具体算法。计算机仿真结果表明该方法简明实用。     

一种基于局部保持的隐变量模型

隐变量模型是一类有效的降维方法,但是由非线性核映射建立的隐变量模型不能保持数据空间的局部结构。为了克服这个缺点,文中提出一种保持数据局部结构的隐变量模型。该算法充分利用局部保持映射的保局性质,将局部保持映射的目标函数作为低维空间中数据的先验信息,对高斯过程隐变量中的低维数据进行约束,建立局部保持的隐变量。实验结果表明,相比原有的高斯过程隐变量,文中算法较好地保持数据局部结构的效果。     

基于时序隐变量模型的因果关系发现算法

为在基于隐变量模型的因果关系发现算法中综合考虑隐变量之间的瞬时性和延时性因果效应,构建以动态贝叶斯网络为基础的时序隐变量模型,提出对应的因果关系发现算法。使用因子分析的方法估计测量模型中的因子载荷矩阵,应用结构向量自回归模型估计自回归矩阵,利用数据的非高斯性依次学习模型中隐变量之间的瞬时效应矩阵与延时效应矩阵,构建时序隐变量模型的因果网络结构。实验结果验证了算法的有效性。     

OpenMP程序中基于活跃变量分析的检查点优化

OpenMP程序通过做检查点来实现容错,现有检查点方法未考虑活跃变量的语义,通过原始值拷贝来保存活跃变量,存在检查点保存量过大的问题。对此提出数组活跃变量检查点优化算法。该算法基于程序分析,处理隐式定值和隐式引用,运用活跃变量分析法得出数组元素中的活跃变量,在适当的情况下采用由下标和数组首地址构成的表达式来表达数组元素,省略其原始值在检查点中的保存,从而减少检查点的数据保存量,达到降低检查点开销的目的。实验表明,该算法可以减少检查点的数据保存量,降低数组元素的数据保存量,达到降低检查点开销的目的,并且消除隐式定值和隐式引用带来的一些不良影响。     

对一种基于IP标识协议隐写的分析

针对一种基于IP标识位协议隐写算法进行了分析,提出可实现隐写分析的统计变量的构造方法,给出分析窗口内含密数据包比率的估计算法,对如何提高算法的安全性给出了改进策略,实验结果表明了该算法的有效性。     

一种任意分布下的隐变量因果结构学习算法

因果发现旨在通过观测数据挖掘变量间的因果关系,在实际应用中需要从观测数据中学习隐变量间的因果结构。现有方法主要利用观测变量间的协方差信息（如四分体约束）或引入非高斯假设（如三分体约束）来解决线性因果模型下的隐变量结构学习问题,但大多限定于分布明确的情况,而实际应用环境往往并不满足这种假设。给出任意分布下隐变量结构的识别性证明,指出在没有混淆因子影响的情况下,两个隐变量的因果方向可识别所需要的最小条件是仅需要其中一个隐变量的噪声服从非高斯分布。在此基础上,针对线性隐变量模型提出一种在任意分布下学习隐变量因果结构的算法,先利用四分体约束方法学习得到隐变量骨架图,再通过枚举骨架图的等价类并测量每一个等价类中的三分体约束来学习因果方向,同时将非高斯约束放宽到尽可能最小的变量子集,从而扩展线性隐变量模型的应用范围。实验结果表明,与MIMBuild和三分体约束方法相比,该算法得到了最佳的F1值,能够在任意分布下学习更多的隐变量因果结构信息,且具有更强的鲁棒性。     

改进的单纯形法迭代计算方法

对传统大M法进行改进,若计算检验数的表达式中含有M则只计算含有M的部分,从而简化计算,迭代过程中当人工变量由基变量变为非基变量时,直接去掉人工变量部分的表格然后继续计算,从而再一次降低计算量。借鉴两阶段法的优点进一步给出了无需给出大M的迭代算法,此法不会破坏目标函数的一致性,而且可以避免传统大M法在利用计算机求解时由于M值的选取不当所导致的计算错误。     

求解SAT问题的线性半定规划算法

将线性半定规划应用到SAT问题的求解过程中。首先将SAT实例转化为整数规划问题,然后松弛为线性规划模型,最后再转化为一般的线性半定规划模型去求解。用SDPA-M软件求解线性半定规划问题后,规定了如何根据目标函数值去判定SAT实例和当CNF公式可满足时如何根据最优指派的概率X^＊i（i=1,…,n）去进行变元赋值,以期求得该公式的可满足指派。上述算法不仅可以判定SAT问题,而且对于符合算法规定可满足的CNF公式皆可给出一个可满足指派。求解SAT问题的线性半定规划算法在文章中被描述并被给予相应算例。     

The second-order cone eigenvalue complementarity problem

The eigenvalue complementarity problem (EiCP) differs from the traditional eigenvalue problem in that the primal and dual variables belong to a closed and convex cone K and its dual, respectively, and satisfy a complementarity condition. In this paper we investigate the solution of the second-order cone EiCP (SOCEiCP) where K is the Lorentz cone. We first show that the SOCEiCP reduces to a special Variational Inequality Problem on a compact set defined by K and a normalization constraint. This guarantees that SOCEiCP has at least one solution, and a new enumerative algorithm is introduced for finding a solution to this problem. The method is based on finding a global minimum of an appropriate nonlinear programming (NLP) formulation of the SOCEiCP using a special branching scheme along with a local nonlinear optimizer that computes stationary points on subsets of the feasible region of NLP associated with the nodes generated by the algorithm. A semi-smooth Newton's method is combined with this enumerative algorithm to enhance its numerical performance. Our computational experience illustrates the efficacy of the proposed techniques in practice.     

A time-optimal control algorithm for digital computer control

A time-optimal control algorithm for digital computer control allowing bounds on control variables and state variables is presented. Through linear programming techniques a time-optimal sequence is computed by using a linear discrete model of the process. For real-time applications feedback control is achieved by recalculating the control sequence each sampling period. In addition, an adaptive control strategy based on on-line estimation of the state variables and the parameters of the system is introduced. In order to be able to apply the algorithm in a real-time environment, computational efficiency is emphasized. An application of the algorithm to a sixth-order multivariable system is given.     

A duality approach for solving bounded linear programming problems with fuzzy variables based on ranking functions and its application in bounded transportation problems

There are several methods, in the literature, for solving fuzzy variable linear programming problems (fuzzy linear programming in which the right-hand-side vectors and decision variables are represented by trapezoidal fuzzy numbers). In this paper, the shortcomings of some existing methods are pointed out and to overcome these shortcomings a new method based on the bounded dual simplex method is proposed to determine the fuzzy optimal solution of that kind of fuzzy variable linear programming problems in which some or all variables are restricted to lie within lower and upper bounds. To illustrate the proposed method, an application example is solved and the obtained results are given. The advantages of the proposed method over existing methods are discussed. Also, one application of this algorithm in solving bounded transportation problems with fuzzy supplies and demands is dealt with. The proposed method is easy to understand and to apply for determining the fuzzy optimal solution of bounded fuzzy variable linear programming problems occurring in real-life situations.     

Large Scale Kernel Regression via Linear Programming

The problem of tolerant data fitting by a nonlinear surface, induced by a kernel-based support vector machine is formulated as a linear program with fewer number of variables than that of other linear programming formulations. A generalization of the linear programming chunking algorithm for arbitrary kernels is implemented for solving problems with very large datasets wherein chunking is performed on both data points and problem variables. The proposed approach tolerates a small error, which is adjusted parametrically, while fitting the given data. This leads to improved fitting of noisy data (over ordinary least error solutions) as demonstrated computationally. Comparative numerical results indicate an average time reduction as high as 26.0% over other formulations, with a maximal time reduction of 79.7%. Additionally, linear programs with as many as 16,000 data points and more than a billion nonzero matrix elements are solved.     

求解整数线性规划问题的定界阻止算法的改进*

在现有求解整数线性规划问题的定界阻止算法的基础上提出了一种改进。该算法通过目标函数超平面截线性规划松弛问题的有效约束锥而形成一个单纯形;然后,引入一串平行片来切割该单纯形产生更低维的凸多面体;最后,在片上的这些凸多面体上执行阻止搜寻程序。由于单纯形和片上凸多面体的极顶点可以直接通过公式计算,且变量在片上凸多面体上的取值区间更窄,改进的定界阻止算法既方便又高效,这得到了一些经典算例和随机产生的算例的验证。     

Multiloop transportation simplex algorithm

In large-scale transportation problems (TPs), various methods have been developed to obtain an optimal solution. One of the methods is the transportation simplex algorithm (TSA), which obtains an optimal solution for TP. Various heuristic methods have been developed to find an initial basic feasible solution for transportation algorithms. These methods differ in terms of computational cost and finding good initial solution. In TSA, the better the basic feasible solution, the less the number of iterations the algorithm will run. At each step, it uses pivoting operation, where a loop involving the nonbasic variable with the largest cost reduction is determined. Then, it eliminates the entering basic variable. However, for large-scale problems, even the best basic feasible solution may result in high number of iterations. In this paper, we present a novel algorithm called multiloop transportation simplex algorithm which finds multiple independent loops during pivoting operation. This causes larger cost reductions in every iteration. We experimentally show that the average number of iterations and runtime to solve the TP are dramatically reduced.     

Integrated staffing and scheduling for an aircraft line maintenance problem

This paper studies the problem of constructing the workforce schedules of an aircraft maintenance company. The problem involves both a staffing and a scheduling decision. We propose an enumerative algorithm with bounding in which each node of the enumeration tree represents a mixed integer linear problem (MILP). We reformulate the MILP such that it becomes tractable for commercial MILP solvers. Extensive computational tests on 40 instances that are derived from a real-life setting indicate that the algorithm is capable of finding close-to-optimal solutions.     

Solving linear programming problems under fuzziness and randomness environment using attainment values

In this paper, the author presents a model to measure attainment values of fuzzy numbers/fuzzy stochastic variables. These new measures are then used to convert the fuzzy linear programming problem or the fuzzy stochastic linear programming problem into the corresponding deterministic linear programming problem. Numerical comparisons are provided to illustrate the effectiveness of the proposed method.