TRUST REGION METHOD IN NEURAL NETWORK

1.IntroductionHopfieldandTank[5]presentedamodeltosolvetravellingsalesmanproblem,thusinitiatingtheapplicationofneuralnetwork(NN)inthefieldofoptimization.SincethenmanyNNmodelshavebeenproposedtosolvelinearprogramming(LP)problems(13,8,11,14,15])andquadraticprogramming(oP)problems([1,8,20]),asLPandoPhavefundamentalimportanceinthetheoryandpracticeofoptimization.Therewerealsoafewmodelsforgeneralnonlinearprogramming(NP)problem([2,6,9,18]).InthispaperwewillpresentaHopfield-typeneuralnetworkmodelw…     

约束非线性规划的神经网络算法

神经网络具有内在大规模并行运算和快速收敛特性,它在最优化技术上的运用近年来受到广泛的重视。本提出一个新的求解一般约束非线性规划的神经网络模型,它具有全局收敛性和广泛的适用性,是求解一般非线性规划问题的新工具。理论分析和模拟计算均表明了模型的有效性。     

Maximally productive input–output units

Technical or Pareto–Koopmans efficiency models can be based on ratios of weighted sums of outputs to weighted sums of inputs. Differing meanings have been considered for such factor weights. In this paper, we use value or cost rate meanings depending on model orientation. These meanings permit considering the simultaneous assignment of input and output factor weights along with optimal intensity values for a virtual composite unit constructed from the observed units. An optimization principle we call the winners-take-all criterion is proposed for determining the maximally productive unit(s). No assumptions are required on the internal transformation processes of the units. The model simultaneously determines the intensities and factor weights and results in indefinite quadratic programming problems that simplify to linear programming in certain cases. For the general case, genetic search is applied. Numerical illustrations are provided for faculty merit scoring and for the 15 hospital dataset of Sherman.     

Neural Approach for Solving Several Types of Optimization Problems

Neural networks consist of highly interconnected and parallel nonlinear processing elements that are shown to be extremely effective in computation. This paper presents an architecture of recurrent neural net-works that can be used to solve several classes of optimization problems. More specifically, a modified Hopfield network is developed and its inter-nal parameters are computed explicitly using the valid-subspace technique. These parameters guarantee the convergence of the network to the equilibrium points, which represent a solution of the problem considered. The problems that can be treated by the proposed approach include combinatorial optimiza-tion problems, dynamic programming problems, and nonlinear optimization problems.Communicated by L. C. W. Dixon     

Group synchronization in complex dynamical networks with different types of oscillators and adaptive coupling schemes

This paper focus on schemes and corresponding criteria for group synchronization in complex dynamical networks consisted of different group of chaotic oscillators. The global asymptotically stable criteria for a linearly or adaptively coupled network are derived to ensure each group of oscillators synchronize to the same behavior. Theoretical analysis and numerical simulation results show that the group synchronization can be guaranteed by enhancing the external coupling strength whenever there are connections or not within the groups under the “same input” condition. All of the results are proved rigorously. Finally, a network with three groups, a scale-free sub-network, a small-world sub-network and a ring sub-network, is illustrated, and the corresponding numerical simulations verify the theoretical analysis.     

一类神经网络模型的稳定性

本文将一种求解凸规划问题的神经网络模型推广到求解一般的非凸非线性规划问题．理论分析表明;在适当的条件下,本文提出的求解非凸非线性规划问题的神经网络模型的平衡点是渐近稳定的,对应于非线性规划问题的局部最优解     

Toward Fuzzy Optimization without Mathematical Ambiguity

Fuzzy programming has been discussed widely in literature and applied in such various disciplines as operations research, economic management, business administration, and engineering. The main purpose of this paper is to present a brief review on fuzzy programming models, and classify them into three broad classes: expected value model, chance-constrained programming and dependent-chance programming. In order to solve general fuzzy programming models, a hybrid intelligent algorithm is also documented. Finally, some related topics are discussed.     

Numerical solution of the nonlinear Schrodinger equation by feedforward neural networks

We present a method to solve boundary value problems using artificial neural networks (ANN). A trial solution of the differential equation is written as a feed-forward neural network containing adjustable parameters (the weights and biases). From the differential equation and its boundary conditions we prepare the energy function which is used in the back-propagation method with momentum term to update the network parameters. We improved energy function of ANN which is derived from Schrodinger equation and the boundary conditions. With this improvement of energy function we can use unsupervised training method in the ANN for solving the equation. Unsupervised training aims to minimize a non-negative energy function. We used the ANN method to solve Schrodinger equation for few quantum systems. Eigenfunctions and energy eigenvalues are calculated. Our numerical results are in agreement with their corresponding analytical solution and show the efficiency of ANN method for solving eigenvalue problems.     

Dynamic multi-criteria evaluation of co-evolution strategies for solving stock trading problems

Risk and return are interdependent in a stock portfolio. To achieve the anticipated return, comparative risk should be considered simultaneously. However, complex investment environments and dynamic change in decision making criteria complicate forecasts of risk and return for various investment objects. Additionally, investors often fail to maximize their profits because of improper capital allocation. Although stock investment involves multi-criteria decision making (MCDM), traditional MCDM theory has two shortfalls: first, it is inappropriate for decisions that evolve with a changing environment; second, weight assignments for various criteria are often oversimplified and inconsistent with actual human thinking processes.In 1965, Rechenberg proposed evolution strategies for solving optimization problems involving real number parameters and addressed several flaws in traditional algorithms, such as their use of point search only and their high probability of falling into optimal solution area. In 1992, Hillis introduced the co-evolutionary concept that the evolution of living creatures is interactive with their environments (multi-criteria) and constantly improves the survivability of their genes, which then expedites evolutionary computation. Therefore, this research aimed to solve multi-criteria decision making problems of stock trading investment by integrating evolutionary strategies into the co-evolutionary criteria evaluation model. Since co-evolution strategies are self-calibrating, criteria evaluation can be based on changes in time and environment. Such changes not only correspond with human decision making patterns (i.e., evaluation of dynamic changes in criteria), but also address the weaknesses of multi-criteria decision making (i.e., simplified assignment of weights for various criteria).Co-evolutionary evolution strategies can identify the optimal capital portfolio and can help investors maximize their returns by optimizing the preoperational allocation of limited capital. This experimental study compared general evolution strategies with artificial neural forecast model, and found that co-evolutionary evolution strategies outperform general evolution strategies and substantially outperform artificial neural forecast models. The co-evolutionary criteria evaluation model avoids the problem of oversimplified adaptive functions adopted by general algorithms and the problem of favoring weights but failing to adaptively adjust to environmental change, which is a major limitation of traditional multi-criteria decision making. Doing so allows adaptation of various criteria in response to changes in various capital allocation chromosomes. Capital allocation chromosomes in the proposed model also adapt to various criteria and evolve in ways that resemble thinking patterns.     

求解线性约束的区间二次规划问题的神经网络

在本文中,基于神经网络,提出了一类求解具有线性约束区间二次规划问题的方法,使用增广拉格朗日函数,建立了求解规划问题的神经网络模型。基于压缩不动点理论,证明了所提出神经网络的平衡点就是等式约束区间二次规划问题的最优解。使用适当的Lyapunov函数,证明了所提出的神经网络的平衡点是全局指数稳定的。最后,两个数值仿真结果验证了本文所用方法的可行性与有效性。     

Wavelet Collocation Method for Optimal Control Problems

A Haar wavelet technique is discussed as a method for discretizing the nonlinear system equations for optimal control problems. The technique is used to transform the state and control variables into nonlinear programming (NLP) parameters at collocation points. A nonlinear programming solver can then be used to solve optimal control problems that are rather general in form. Here, general Bolza optimal control problems with state and control constraints are considered. Examples of two kinds of optimal control problems, continuous and discrete, are solved. The results are compared to those obtained by using other collocation methods.     

Inferring Electre’s veto-related parameters from outranking examples

When considering Electre’s valued outranking relations, aggregation/disaggregation methodologies have difficulties in taking discordance (veto) into account. We present a partial inference procedure to compute the value of the veto-related parameters that best restore a set of outranking statements (i.e., examples that an Electre model should restore) provided by a decision maker, given fixed values for the remaining parameters of the model.This paper complements previous work on the inference of other preference-related parameters (weights, cutting level, category limits, …), advancing toward an integrated framework of inference problems in Electre III and Tri methods. We propose mathematical programs to infer veto-related parameters, first considering only one criterion, then all criteria simultaneously, using the original version of Electre outranking relation and two variants. This paper shows that these inference procedures lead to linear programming, 0–1 linear programming, or separable programming problems, depending on the case.     

A tolerance approach to the fuzzy goal programming problems with unbalanced triangular membership function

This paper investigates the application of tolerance concepts to goal programming in a fuzzy environment. Firstly, the paper presents how the fuzzy goal programming (FGP) problems with unequal weights can be formulated as a single linear programming problem with the concept of tolerances. Next, it illustrates a numerical example to demonstrate how convenient the present model is to solve a FGP problem with unequal weights and unbalanced linear membership functions     

Integer Programming Duality in Multiple Objective Programming

The weighted sums approach for linear and convex multiple criteria optimization is well studied. The weights determine a linear function of the criteria approximating a decision makers overall utility. Any efficient solution may be found in this way. This is not the case for multiple criteria integer programming. However, in this case one may apply the more general e-constraint approach, resulting in particular single-criteria integer programming problems to generate efficient solutions. We show how this approach implies a more general, composite utility function of the criteria yielding a unified treatment of multiple criteria optimization with and without integrality constraints. Moreover, any efficient solution can be found using appropriate composite functions. The functions may be generated by the classical solution methods such as cutting plane and branch and bound algorithms.     

On min-norm and min-max methods of multi-objective optimization

This paper examines Pareto optimality of solutions to multi-objective problems scalarized in the min-norm, compromise programming, generalized goal programming, or unrestricted min-max formulations. Issues addressed include, among others, uniqueness in solution or objective space, penalization for over-achievement of goals, min-max reformulation of goal programming, inferiority in Tchebycheff-norm minimization, strength and weakness of weighted-bound optimization, quasi-satisficing decision-making, just attaining or even over-passing the goals, trading off by modifying weights or goals, non-convex Pareto frontier. New general necessary and sufficient conditions for both Pareto optimality and weak Pareto optimality are presented. Various formulations are compared in theoretical performance with respect to the goal-point location. Ideas for advanced goal programming and interactive decision-making are introduced.     

部分权重信息下对方案有互补偏好的多属性决策法

研究了只有部分权重信息且对方案的偏好信息以模糊互补判断矩阵形式给出的多属性决策问题.首先,基于模糊互补判断矩阵的主观偏好信息,利用转换函数将客观决策信息一致化,建立一个目标规划模型,通过求解该模型得到属性权重,从而利用加性加权法获得各方案的综合属性值,并以此对方案进行排序或择优.提出了一种基于目标规划的多属性决策方法.该方法具有操作简便和易于上机实现的特点.最后,通过实例说明模型及方法的可行性和有效性.     

Note on “Some models for deriving the priority weights from interval fuzzy preference relations”

Deriving accurate interval weights from interval fuzzy preference relations is key to successfully solving decision making problems. Xu and Chen (2008) proposed a number of linear programming models to derive interval weights, but the definitions for the additive consistent interval fuzzy preference relation and the linear programming model still need to be improved. In this paper, a numerical example is given to show how these definitions and models can be improved to increase accuracy. A new additive consistency definition for interval fuzzy preference relations is proposed and novel linear programming models are established to demonstrate the generation of interval weights from an interval fuzzy preference relation.     

Continuous-Time Generalized Fractional Programming Problems. Part I: Basic Theory

This study, that will be presented as two parts, develops a computational approach to a class of continuous-time generalized fractional programming problems. The parametric method for finite-dimensional generalized fractional programming is extended to problems posed in function spaces. The developed method is a hybrid of the parametric method and discretization approach. In this paper (Part I), some properties of continuous-time optimization problems in parametric form pertaining to continuous-time generalized fractional programming problems are derived. These properties make it possible to develop a computational procedure for continuous-time generalized fractional programming problems. However, it is notoriously difficult to find the exact solutions of continuous-time optimization problems. In the accompanying paper (Part II), a further computational procedure with approximation will be proposed. This procedure will yield bounds on errors introduced by the numerical approximation. In addition, both the size of discretization and the precision of an approximation approach depend on predefined parameters.     

A multi-parametric programming approach for multilevel hierarchical and decentralised optimisation problems

In this paper, we outline the foundations of a general global optimisation strategy for the solution of multilevel hierarchical and general decentralised multilevel problems, based on our recent developments on multi-parametric programming and control theory. The core idea is to recast each optimisation subproblem, present in the hierarchy, as a multi-parametric programming problem, with parameters being the optimisation variables belonging to the remaining subproblems. This then transforms the multilevel problem into single-level linear/convex optimisation problems. For decentralised systems, where more than one optimisation problem is present at each level of the hierarchy, Nash equilibrium is considered. A three person dynamic optimisation problem is presented to illustrate the mathematical developments.     

Vector and Parametric Programming

Vector programming seeks to identify efficient solutions of multiple-objective problems. It is closely related to parametric programming and three examples of this relationship, using objective weights, target (or goal) levels and weighted norms, are described. In each case, the use of the relationship to develop algorithms for determining the efficient set or to obtain characterizations of efficiency is investigated.