Stability in multiobjective possibilistic linear programs

Similar documents
On stability in possibilistic linear equality systems with Lipschitzian fuzzy numbers

On Fuzzy Internal Rate of Return

AN EFFICIENT APPROACH FOR SOLVING A WIDE CLASS OF FUZZY LINEAR PROGRAMMING PROBLEMS

Numerical Solution of Fuzzy Differential Equations

Institute for Advanced Management Systems Research Department of Information Technologies Åbo Akademi University. Fuzzy Logic Controllers - Tutorial

A Method for Solving Fuzzy Differential Equations Using Runge-Kutta Method with Harmonic Mean of Three Quantities

On additions of interactive fuzzy numbers

EXTRACTING FUZZY IF-THEN RULE BY USING THE INFORMATION MATRIX TECHNIQUE WITH QUASI-TRIANGULAR FUZZY NUMBERS

Fuzzy if-then rules for modeling interdependences in FMOP problems

Fuzzy Multi-objective Linear Programming Problem Using Fuzzy Programming Model

NUMERICAL SOLUTIONS OF FUZZY DIFFERENTIAL EQUATIONS BY TAYLOR METHOD

Special Classes of Fuzzy Integer Programming Models with All-Dierent Constraints

Chi-square goodness-of-fit test for vague data

A Linear Regression Model for Nonlinear Fuzzy Data

A METHOD FOR SOLVING FUZZY PARTIAL DIFFERENTIAL EQUATION BY FUZZY SEPARATION VARIABLE

Multiattribute decision making models and methods using intuitionistic fuzzy sets

Previous Accomplishments. Focus of Research Iona College. Focus of Research Iona College. Publication List Iona College. Journals

Research Article A Compensatory Approach to Multiobjective Linear Transportation Problem with Fuzzy Cost Coefficients

On fuzzy entropy and topological entropy of fuzzy extensions of dynamical systems

Using ranking functions in multiobjective fuzzy linear programming 1

FIXED POINT THEOREMS OF KRASNOSELSKII TYPE IN A SPACE OF CONTINUOUS FUNCTIONS

Necessary and Sufficient Optimality Conditions for Nonlinear Fuzzy Optimization Problem

SOLVING FUZZY DIFFERENTIAL EQUATIONS BY USING PICARD METHOD

Two Step Method for Fuzzy Differential Equations

A pure probabilistic interpretation of possibilistic expected value, variance, covariance and correlation

On the Continuity and Convexity Analysis of the Expected Value Function of a Fuzzy Mapping

FUZZY SOLUTIONS FOR BOUNDARY VALUE PROBLEMS WITH INTEGRAL BOUNDARY CONDITIONS. 1. Introduction

Fuzzy Reasoning and Fuzzy Optimization

On Informational Coefficient of Correlation for Possibility Distributions

Downloaded from iors.ir at 10: on Saturday May 12th 2018 Fuzzy Primal Simplex Algorithms for Solving Fuzzy Linear Programming Problems

A possibilistic approach to risk premium

Differential Equations based on Fuzzy Rules

Weak convergence and large deviation theory

Solution of Fuzzy Maximal Flow Network Problem Based on Generalized Trapezoidal Fuzzy Numbers with Rank and Mode

UC Berkeley Department of Electrical Engineering and Computer Science. EECS 227A Nonlinear and Convex Optimization. Solutions 5 Fall 2009

On the levels of fuzzy mappings and applications to optimization

GENERATING LARGE INDECOMPOSABLE CONTINUA

Key Renewal Theory for T -iid Random Fuzzy Variables

On Measures of Dependence Between Possibility Distributions

Chapter 1 Similarity Based Reasoning Fuzzy Systems and Universal Approximation

Solving Dual Problems

Evaluation of Fuzzy Linear Regression Models by Parametric Distance

IN many real-life situations we come across problems with

A MAXIMUM PRINCIPLE FOR A MULTIOBJECTIVE OPTIMAL CONTROL PROBLEM

On Weak Pareto Optimality for Pseudoconvex Nonsmooth Multiobjective Optimization Problems

Hausdorff Continuous Viscosity Solutions of Hamilton-Jacobi Equations

Prediction-based adaptive control of a class of discrete-time nonlinear systems with nonlinear growth rate

On possibilistic correlation

PAijpam.eu OBTAINING A COMPROMISE SOLUTION OF A MULTI OBJECTIVE FIXED CHARGE PROBLEM IN A FUZZY ENVIRONMENT

Duality in fuzzy linear programming: A survey

Lagrangian Duality Theory

Fuzzy Mathematical Programming: Theory, Applications and Extension

Tuning of Fuzzy Systems as an Ill-Posed Problem

"SYMMETRIC" PRIMAL-DUAL PAIR

Some chaotic and mixing properties of Zadeh s extension

Research Article A Fictitious Play Algorithm for Matrix Games with Fuzzy Payoffs

DIFFERENCE METHODS FOR FUZZY PARTIAL DIFFERENTIAL EQUATIONS

Numerical Solving of a Boundary Value Problem for Fuzzy Differential Equations

Fuzzy Sets and Fuzzy Techniques. Sladoje. Introduction. Operations. Combinations. Aggregation Operations. An Application: Fuzzy Morphologies

Section Notes 9. Midterm 2 Review. Applied Math / Engineering Sciences 121. Week of December 3, 2018

Numerical Method for Solving Fuzzy Nonlinear Equations

Introduction to Intelligent Control Part 6

Existence and data dependence for multivalued weakly Ćirić-contractive operators

CREDIBILITY THEORY ORIENTED PREFERENCE INDEX FOR RANKING FUZZY NUMBERS

Structural Design under Fuzzy Randomness

Conditioning of linear-quadratic two-stage stochastic programming problems

A Note on the Class of Superreflexive Almost Transitive Banach Spaces

Fuzzy Order Statistics based on α pessimistic

Application of the Fuzzy Weighted Average of Fuzzy Numbers in Decision Making Models

From now on, we will represent a metric space with (X, d). Here are some examples: i=1 (x i y i ) p ) 1 p, p 1.

Statistical Hypotheses Testing in the Fuzzy Environment

Converse Lyapunov theorem and Input-to-State Stability

Interactive Random Fuzzy Two-Level Programming through Possibility-based Fractile Criterion Optimality

Optimization Problems with Probabilistic Constraints

The local equicontinuity of a maximal monotone operator

SOLVING FUZZY LINEAR SYSTEMS OF EQUATIONS

Fuzzy Distance Measure for Fuzzy Numbers

Extremal Solutions of Differential Inclusions via Baire Category: a Dual Approach

A note on the products of the terms of linear recurrences

/07/$ IEEE

EE/ACM Applications of Convex Optimization in Signal Processing and Communications Lecture 17

Institute for Statics und Dynamics of Structures Fuzzy probabilistic safety assessment

Introduction to Nonlinear Stochastic Programming

An Evaluation of the Reliability of Complex Systems Using Shadowed Sets and Fuzzy Lifetime Data

On intermediate value theorem in ordered Banach spaces for noncompact and discontinuous mappings

Solving fuzzy matrix games through a ranking value function method

An Implicit Method for Solving Fuzzy Partial Differential Equation with Nonlocal Boundary Conditions

Fuzzy Systems. Fuzzy Control

Uncertain Systems are Universal Approximators

Quiz Discussion. IE417: Nonlinear Programming: Lecture 12. Motivation. Why do we care? Jeff Linderoth. 16th March 2006

Convergence rate estimates for the gradient differential inclusion

Solution of the Fuzzy Boundary Value Differential Equations Under Generalized Differentiability By Shooting Method

Nonlinear Optimization Subject to a System of Fuzzy Relational Equations with Max-min Composition

Linear Programming Methods

A CHARACTERIZATION OF STRICT LOCAL MINIMIZERS OF ORDER ONE FOR STATIC MINMAX PROBLEMS IN THE PARAMETRIC CONSTRAINT CASE

Stability of efficient solutions for semi-infinite vector optimization problems

Sample width for multi-category classifiers

2 Statement of the problem and assumptions

The Subdifferential of Convex Deviation Measures and Risk Functions

Section 45. Compactness in Metric Spaces

Transcription:

Stability in multiobjective possibilistic linear programs Robert Fullér rfuller@abo.fi Mario Fedrizzi fedrizzi@cs.unitn.it Abstract This paper continues the authors research in stability analysis in possibilistic programming in that it extends the results in [7] to possibilistic linear programs with multiple objective functions. Namely, we show that multiobjective possibilistic linear programs with continuous fuzzy number coefficients are well-posed, i.e. small changes in the membership function of the coefficients may cause only a small deviation in the possibility distribution of the objective function. Keywords: Multiobjective possibilistic linear programs (MPLP), possibility theory, fuzzy number, stability 1 Introduction Stability and sensitivity analysis becomes more and more attractive also in the area of multiple objective mathematical programming (for excellent surveys see e.g. Gal [10] and Rios Insua [18]). Publications on this topic usually investigate the impact of parameter changes (in the righthand side or/and the objective functions or/and the A-matrix or/and the domination structure) on the solution in various models of vectormaximization problems, e.g. linear or non-linear, deterministic or stochastic, static or dynamic (see e.g. [5, 17, 19]). There are only few paper dealing with stability or sensitivity analysis in fuzzy mathematical programming: Sensitivity analysis in LP problems with crisp coefficients and soft constraints was first considered in [11], where a functional relationship between changes of parameters of the righthand side and those of the optimal value of the primal objective function was derived for almost all conceivable cases. The stability of the fuzzy solution in LP problems with fuzzy coefficients of symmetric triangular form was shown in [9]. In this paper, generalizing the authors earlier result [7], we show that the possibility distribution of the objectives of an MPLP with continuous fuzzy number coefficients is stable under small changes in the membership function of the parameters. In this section we recall Buckley s [2, 3, 4] and The final version of this paper appeared in: R. Fullér and M. Fedrizzi, Stability in multiobjective possibilistic linear programs, European Journal of Operational Research, 74(1994) 179-187. 1

Luhandjula s [16] solution concept for MPLP and set up the notation needed for our main result in the next section. A fuzzy quantity ã: IR [0, 1] is a fuzzy set of the real line IR. The support of a fuzzy quantity ã (denoted by suppã) is the crisp set given by {t ã(t) > 0}. A fuzzy number ã is a fuzzy quantity with a normalized, upper semicontinuous, fuzzy convex and bounded supported membership function. A symmetric triangular fuzzy number ã denoted by (a, α) is defined as a t 1 if a t α, ã(t) = α where a IR is the centre and 2α > 0 is the spread of ã. The fuzzy numbers will represent the continuous possibility distributions for fuzzy parameters. A multiobjective possibilistic linear programming is maximize Z = ( c 11 x 1 + + c 1n x n,..., c k1 x 1 + + c kn x n ) (1) subject to ã i1 x 1 + ã in x n b i, i = 1,..., m, x 0, where ã ij, b i, and c lj are fuzzy numbers, x = (x 1,..., x n ) is a vector of (non-fuzzy) decision variables, the operations addition and multiplication by a real number of fuzzy numbers are defined by Zadeh s extension principle [20], and denotes <,, =, or > for each i, i = 1,..., m. Even though may vary from row to row in the constraints, we will rewrite the MPLP (1) as maximize Z = ( c 1 x,..., c k x) (2) subject to Ãx b, x 0, where ã = [ã ij ] is an m n matrix of fuzzy numbers and b = ( b 1,..., b m ) is a vector of fuzzy numbers. The fuzzy numbers are the possibility distributions associated with the fuzzy variables and hence place a restriction on the possible values the variable may assume [20, 21]. For example, Poss[ã ij = t] = ã ij (t). We will assume that all fuzzy numbers ã ij, b i, c j are non-interactive [20]. Following Buckley [2-4], we define Poss[Z = z], the possibility distribution of the objective function Z. We first specify the possibility that x satisfies the i-th constraints. Let π(a i, b i ) = min{ã i1 (a i1 ),..., ã in (a in ), b i (b i }, where a i = (a i1,..., a in ), which is the joint distribution of ã ij, j = 1,..., n, and b i. Then Poss[x F i ] = sup a i,b i { π(a i, b i ) a i1 x 1 +... + a in x n b i }, which is the possibility that x is feasible with respect to the i-th constraint. Therefore, for x 0, Poss[x F] = min 1 i m Poss[x F i], which is the possibility that x is feasible. We next construct Poss[Z = z x] which is the conditional possibility that Z equals z given x. The joint distribution of the c lj, j = 1,..., n, is π(c l ) = min{ c l1 (c l1 ),..., c ln (c ln )} 2

where c l = (c l1,..., c ln ), l = 1,..., k. Therefore, Poss[Z = z x] = Poss[ c 1 x = z 1,..., c k x = z k ] = min 1 l k Poss[ c lx = z l ] = min 1 l k sup {π(c l ) c l1 x 1 + + c ln x n = z l }. c l1,...,c lk Finally, applying Bellman and Zadeh s method of fuzzy decision making [1], the possibility distribution of the objective function is defined as follows Poss[Z = z] = sup min{poss[z = z x], Poss[x F]} x 0 Let ã be a fuzzy number. Then for any θ 0 we define ω(ã, θ), the modulus of continuity of ã as [12] ω(ã, θ) = max ã(u) ã(v). u v θ An α-level set of a fuzzy numer ã is a non-fuzzy set denoted by [ã] α and is defined by { {t IR ã(t) α} if α > 0 [ã] α = cl(suppã) if α = 0 where cl(suppã) denotes the closure of the support of ã. Let ã and b be fuzzy numbers and [ã] α = [a 1 (α), a 2 (α)], [ b] α = [b 1 (α), b 2 (α)]. We metricize the set of fuzzy numbers by the metric [13] D(ã, b) = sup max { a i(α) b i (α) }, α [0,1] i=1,2 where d denotes the classical Hausdorff metric in the family of compact subsets of IR 2, i.e. d([ã] α, [ b] α ) = max{ a 1 (α) b 1 (α), a 2 (α) b 2 (α) } The proof of the next two lemmas is very simple. Lemma 1.1 Let n N and let x i, y i [0, 1], i = 1,..., n. Then min i=1,...,n x i min i=1,...,n y i max i=1,...,n x i y i Lemma 1.2 Let X be a set and let f, g be fuzzy sets of X. Then sup x X f(x) sup g(x) sup f(x) g(x) x X x X The following lemma shows that if all the α-level sets of two continuous fuzzy numbers are close to each other, then there can be only a small deviation between their membership grades. Lemma 1.3 [7] Let δ 0 and let ã, b be fuzzy numbers. If D(ã, b) δ, then sup ã(t) b(t) max{ω(ã, δ), ω( b, δ)}. t IR Remark 1.1 It should be noted that if ã or b are discontinuous fuzzy numbers, then there can be a big deviation between their membership grades even if D(ã, b) is arbitrarily small. 3

2 Stability in multiobjective possibilistic linear programs An important question [6, 7, 9, 22] is the influence of the perturbations of the fuzzy parameters to the Possibility distribution of the objective function. We will assume that there is a collection of fuzzy parameters ã δ ij, b δ i, c δ j available with the property max D(ã ij, ã δ ij) δ, i,j max D( b i, b δ i ) δ, max D( c lj, c δ i lj) δ. (3) l,j Then we have to solve the following problem: max/min Z δ = ( c δ 1x,... c δ kx) (4) subject to Ãδ x b δ, x 0. Let us denote by Poss[x F δ i ] the Possibility that x is feasible with respect to the i-th constraint in (4). Then the Possibility distribution of the objective function Z δ in (4) is defined as follows: Poss[Z δ = z] = sup(min{poss[z δ = z x], Poss[x F δ ]}). x 0 The following theorem shows a stability property (with respect to perturbations (3)) of the possibility dostribution of the objective function of MPLP (1) and (4). Theorem 2.1 Let δ 0 be a real number and let ã ij, b i, ã δ ij, c lj, c δ lj fuzzy numbers. If (3) hold, then be continuous sup Poss[Z δ = z] Poss[Z = z] ω(δ) (5) z IR where ω(δ) = max i,j,l {ω(ã ij, δ), ω(ã δ ij, δ), ω( b i, δ), ω(b δ i, δ), ω( c lj, δ), ω( c δ lj, δ)}. i.e. ω(δ) denotes the maximum of modulus of continuity of all fuzzy coefficients at the point δ. Proof. It is sufficient to show that Poss[Z = z] Poss[Z δ = z] ω(δ), for any (z 1,..., z k ) IR k, (6) Applying Lemmas 1.1-1.3 we have... (for full proof see EJOR, 74(1994) 179-187)... which proves the theorem. Remark 2.1 From lim δ 0 ω(δ) = 0 and (5) it follows that sup Poss[Z δ = z] Poss[Z = z] 0 as δ 0 z IR which means the stability of the possibility distribution of the objective function with respect to perturbations (3). 4

Remark 2.2 As an immediate consequence of this theorem we obtain the following result (see [9]: If the fuzzy numbers in (1) and (4) satisfy the Lipschitz condition with constant L > 0, then sup Poss[Z δ = z] Poss[Z = z] Lδ z IR Furthermore, similar estimations can be obtained in the case of symmetrical trapezoidal fuzzy number parameters (see [15]) and in the case of symmetrical triangular fuzzy number parameters (see [8, 14]). Remark 2.3 It is easy to see that in the case of non-continuous fuzzy parameters the possibility distribution of the objective function may be unstable under small changes of the parameters. 3 Illustration As an example, consider the following biobjective possibilistic linear program max/min ( cx, cx) (7) subject to ãx b, x 0. where ã = (1, 1), b = (2, 1) and c = (3, 1) are fuzzy numbers of symmetric triangular form. Here x is one-dimensional (n = 1) and there is only one constraint (m = 1). We find 1 if x 2, Pos[x F] = 3 if x > 2. x + 1 and Pos[Z = (z 1, z 2 ) x] = min{pos[ cx = z 1 ], Pos[ cx = z 2 ]} where for i = 1, 2 and x 0, and Pos[ cx = z i ] = 4 z i x if z i/x [3, 4], z i x 2 if z i/x [2, 3], Pos[Z = (z 1, z 2 ) 0] = Pos[0 c = z] = { 1 if z = 0, 0 otherwise. Both possibilities are nonlinear functions of x, however the calculation of Pos[Z = (z 1, z 2 )] is easily performed and we obtain θ 1 if z M 1, Pos[Z = (z 1, z 2 )] = min{θ 1, θ 2, θ 3 } if z M 2, 5

where and for i = 1, 2 and Figure 1: Pos[ cx = z 2 ] for z 2 = 8. M 1 = {z IR 2 z 1 z 2 min{z 1, z 2 }, z 1 + z 2 12}, M 2 = {z IR 2 z 1 z 2 min{z 1, z 2 }, z 1 + z 2 > 12}, θ i = 24 z i + 7 + z 2 i + 14z i + 1. θ 3 = 4 min{z 1, z 2 } 2 max{z 1, z 2 } z 1 + z 2. Consider now a perturbed biobjective problem with two different objectives (derived from (7) by a simple δ-shifting of the centres of ã and c): max/min ( cx, c δ x) (8) subject to ã δ x b, x 0. where ã = (1 + δ, 1), b = (2, 1), c = (3, 1), c δ = (3 δ, 1) and δ 0 is the error of measurement. Then 1 if x 2 Pos[x F δ ] = 1 + δ, 3 δx if x > 2 x + 1 1 + δ. and Pos[Z δ = (z 1, z 2 ) x] = min{pos[ cx = z 1 ], Pos[ c δ x = z 2 ]} where Pos[ cx = z 1 ] = 4 z 1 x if z 1/x [3, 4], z 1 x 2 if z 1i/x [2, 3], 4 δ z 2 x if z 2/x [3 δ, 4 δ], Pos[ c δ x = z 2 ] = z 2 x 2 + δ if z 2/x [2 δ, 3 δ], 6

x 0, and So, where and M 1 (δ) = M 2 (δ) = Pos[Z δ = (z 1, z 2 ) 0] = Pos[0 c = z] = { 1 if z = 0, 0 otherwise. θ 1 (δ) if z M 1 (δ), Pos[Z δ = (z 1, z 2 )] = min{θ 1 (δ), θ 2 (δ), θ 3 (δ)} if z M 2 (δ), { { z IR 2 z 1 z 2 (1 0.5δ) min{z 1, z 2 }, z 1 + z 2 z IR 2 z 1 z 2 (1 0.5δ) min{z 1, z 2 }, z 1 + z 2 > ( 24 + δ 7 z 1 ) z1 2 + 14z 1 + 1 + 4z 1 δ θ 1 (δ) = z 1 + 7 + z1 2 + 14z 1 + 1 + 4z 1 δ + 2δ 24 δ (δ + z 2 1 + ) (1 δ z 2 ) 2 + 16z 2 θ 2 (δ) = z 2 + 7 + (1 δ z 2 ) 2 + 16z 2 + δ θ 3 (δ) = (4 δ) min{z 1, z 2 } 2 max{z 1, z 2 } z 1 + z 2. It is easy to check that sup Pos[x F] Pos[x F δ ] δ, x 0 sup Pos[Z = z x] Pos[Z δ = z x] δ, x 0, z } 2(6 δ), 1 + δ } 2(6 δ), 1 + δ sup Pos[Z = z] Pos[Z δ = z] δ. z On the other hand, from the definition of metric D the modulus of continuity and Theorem 2.1 it follows that D(ã, ã δ ) = δ, D( c, c δ ) = δ, D( c, c) = 0, D( b, b) = 0, ω(δ) = δ, and, therefore, sup Pos[Z = z] Pos[Z δ = z] δ. z 7

References [1] R.A.Bellman and L.A.Zadeh, Decision-making in a fuzzy environment, Management Sciences, Ser. B 17 (1970) 141-164. [2] J.J.Buckley, Possibilistic linear programming with triangular fuzzy numbers, Fuzzy sets and Systems, 26(1988) 135-138. [3] J.J.Buckley, Solving possibilistic linear programming problems, Fuzzy sets and Systems, 31(1989) 329-341. [4] J.J.Buckley, Multiobjective possibilistic linear programming, Fuzzy Sets and Systems, 35(1990) 23-28. [5] D.V.Deshpande, S.Zionts, Sensitivity analysis in multiply objective linear programming: Changes in the objective function matrix, Working paper no. 399(1979), State University of New York, USA. [6] D.Dubois, Linear programming with fuzzy data, in: J.C.Bezdek ed., Analysis of fuzzy information, Vol.3: Applications in Engineering and Science,CRC Press, Boca Raton, FL,1987. [7] M.Fedrizzi and R.Fullér, Stability in possibilistic linear programming problems with continuous fuzzy number parameters, Fuzzy Sets and Systems, 47(1992) 187-1991. [8] R.Fullér, On stability in fuzzy linear programming problems, Fuzzy Sets and Systems, 30(1989) 339-344. [9] R.Fullér, Stability in possibilistic linear equality systems with Lipschitzian fuzzy numbers, Fuzzy Sets and Systems, 34(1990) 347-353. [10] T.Gal and K.Wolf, Stability in vector maximization - A survey,european Journal of Operational Research, 25(1986) 169-182. [11] H. Hamacher, H. Leberling and H.-J. Zimmermann, Sensitivity analysis in fuzzy linear programming, Fuzzy Sets and Systems 1 (1978) 269-281. [12] P.Henrici, Discrete Variable Methods in Ordinary Differential Equations, John Wiley & Sons, New York, 1962. [13] O.Kaleva, Fuzzy differential equations, Fuzzy Sets and Systems, 24(1987) 301-317. [14] M.Kovács, Fuzzification of ill-posed linear systems, in: D. Greenspan and P. Rózsa, Eds., Colloquia Mathematica Societias János Bolyai 50: Numerical Methods (North-Holland, Amsterdam, 1988) 521-532. [15] M.Kovács,F.P.Vasil jev and R.Fullér, On stability in fuzzified linear equality systems, Proceedings of Moscow State University, 1(1989), Ser.15, 5-9. 8

[16] M.K.Luhandjula, Multiple objective programming problems with possibilistic coefficients, Fuzzy Sets and Systems, 21(1987) 135-145. [17] H.M.Rarig and Y.Y.Haimes, Risk/Dispersion index method, IEEE Transactions on systems, Man, and Cybernetics, 13(1983) 317-328. [18] D.Rios Insua, Sensitivity Analysis in Multi-Objective Decision Making, Springer-Verlag, Berlin, 1990. [19] D.Rios Insua and S.French, A framework for sensitivity analysis in discrete multi-objective decision-making, European Journal of Operational Research, 54(1991) 176-190. [20] L.A.Zadeh, The concept of linguistic variable and its applications to approximate reasoning, Parts I,II,III, Information Sciences, 8(1975) 199-251; 8(1975) 301-357; 9(1975) 43-80. [21] L.A.Zadeh, Fuzzy sets as a basis for theory of possibility, Memo UCB/ERL M77/12, Univ. of California, Berkeley, 1977. [22] H.-J.Zimmermann, Applications of fuzzy set theory to mathematical programming, Information Sciences, 36(1985) 29-58. 9