Approximate Reasoning for Solving Fuzzy Linear Programming Problems



Similar documents
Optimization under fuzzy if-then rules

A Method for Solving Linear Programming Problems with Fuzzy Parameters Based on Multiobjective Linear Programming Technique

What is Linear Programming?

Proximal mapping via network optimization

Optimization of Fuzzy Inventory Models under Fuzzy Demand and Fuzzy Lead Time

No: Bilkent University. Monotonic Extension. Farhad Husseinov. Discussion Papers. Department of Economics

Fuzzy Probability Distributions in Bayesian Analysis

Nonlinear Programming Methods.S2 Quadratic Programming

Neuro-Fuzzy Methods. Robert Fullér Eötvös Loránd University, Budapest. VACATION SCHOOL Neuro-Fuzzy Methods for Modelling & Fault Diagnosis

A New Method for Multi-Objective Linear Programming Models with Fuzzy Random Variables

Portfolio selection based on upper and lower exponential possibility distributions

Stochastic Inventory Control

Linear Programming Notes V Problem Transformations

We shall turn our attention to solving linear systems of equations. Ax = b

MATH 304 Linear Algebra Lecture 18: Rank and nullity of a matrix.

Fuzzy Numbers in the Credit Rating of Enterprise Financial Condition

Product Selection in Internet Business, A Fuzzy Approach

Fuzzy regression model with fuzzy input and output data for manpower forecasting

1 Introduction. Linear Programming. Questions. A general optimization problem is of the form: choose x to. max f(x) subject to x S. where.

Optimal shift scheduling with a global service level constraint

Duality in General Programs. Ryan Tibshirani Convex Optimization /36-725

Some Research Problems in Uncertainty Theory

LECTURE: INTRO TO LINEAR PROGRAMMING AND THE SIMPLEX METHOD, KEVIN ROSS MARCH 31, 2005

7 Gaussian Elimination and LU Factorization

Chapter 3: Section 3-3 Solutions of Linear Programming Problems

Linear Equations in One Variable

FUZZY CLUSTERING ANALYSIS OF DATA MINING: APPLICATION TO AN ACCIDENT MINING SYSTEM

4.6 Linear Programming duality

A Robust Formulation of the Uncertain Set Covering Problem

A FUZZY LOGIC APPROACH FOR SALES FORECASTING

Lecture 13 Linear quadratic Lyapunov theory

Least-Squares Intersection of Lines

2.3 Convex Constrained Optimization Problems

JUST THE MATHS UNIT NUMBER 1.8. ALGEBRA 8 (Polynomials) A.J.Hobson

Support Vector Machines Explained

A New Interpretation of Information Rate

Linear Programming I

Equations, Inequalities & Partial Fractions

Zeros of a Polynomial Function

Lecture 3. Linear Programming. 3B1B Optimization Michaelmas 2015 A. Zisserman. Extreme solutions. Simplex method. Interior point method

Numerisches Rechnen. (für Informatiker) M. Grepl J. Berger & J.T. Frings. Institut für Geometrie und Praktische Mathematik RWTH Aachen

Fuzzy Differential Systems and the New Concept of Stability

On Quantum Hamming Bound

Several Views of Support Vector Machines

Lecture 11: 0-1 Quadratic Program and Lower Bounds

1.5. Factorisation. Introduction. Prerequisites. Learning Outcomes. Learning Style

SOLVING LINEAR SYSTEMS

TOPIC 4: DERIVATIVES

Date: April 12, Contents

Approximation Algorithms

Inner Product Spaces

Duality of linear conic problems

Factoring Cubic Polynomials

Practical Guide to the Simplex Method of Linear Programming

Mathematical finance and linear programming (optimization)

Methods for Finding Bases

Applied Algorithm Design Lecture 5

24. The Branch and Bound Method

International Doctoral School Algorithmic Decision Theory: MCDA and MOO

MATRIX ALGEBRA AND SYSTEMS OF EQUATIONS

SECOND DERIVATIVE TEST FOR CONSTRAINED EXTREMA

Vector and Matrix Norms

Research Article Stability Analysis for Higher-Order Adjacent Derivative in Parametrized Vector Optimization

Lecture 1: Systems of Linear Equations

Solving Linear Systems, Continued and The Inverse of a Matrix

Optimization in R n Introduction

High-Mix Low-Volume Flow Shop Manufacturing System Scheduling

a 11 x 1 + a 12 x a 1n x n = b 1 a 21 x 1 + a 22 x a 2n x n = b 2.

Application. Outline. 3-1 Polynomial Functions 3-2 Finding Rational Zeros of. Polynomial. 3-3 Approximating Real Zeros of.

Linear programming with fuzzy parameters: An interactive method resolution q

MATRIX ALGEBRA AND SYSTEMS OF EQUATIONS. + + x 2. x n. a 11 a 12 a 1n b 1 a 21 a 22 a 2n b 2 a 31 a 32 a 3n b 3. a m1 a m2 a mn b m

The Method of Partial Fractions Math 121 Calculus II Spring 2015

A Log-Robust Optimization Approach to Portfolio Management

5 INTEGER LINEAR PROGRAMMING (ILP) E. Amaldi Fondamenti di R.O. Politecnico di Milano 1

Linear Equations and Inequalities

Nonlinear Optimization: Algorithms 3: Interior-point methods

Largest Fixed-Aspect, Axis-Aligned Rectangle

1 Lecture: Integration of rational functions by decomposition

Integrating Benders decomposition within Constraint Programming

ON LIMIT LAWS FOR CENTRAL ORDER STATISTICS UNDER POWER NORMALIZATION. E. I. Pancheva, A. Gacovska-Barandovska

4. Continuous Random Variables, the Pareto and Normal Distributions

PUTNAM TRAINING POLYNOMIALS. Exercises 1. Find a polynomial with integral coefficients whose zeros include

Chapter 6. Linear Programming: The Simplex Method. Introduction to the Big M Method. Section 4 Maximization and Minimization with Problem Constraints

Introduction to Support Vector Machines. Colin Campbell, Bristol University

3.6. Partial Fractions. Introduction. Prerequisites. Learning Outcomes

Algebra 2 Chapter 1 Vocabulary. identity - A statement that equates two equivalent expressions.

On the representability of the bi-uniform matroid

Introduction to Linear Programming (LP) Mathematical Programming (MP) Concept

Recovery of primal solutions from dual subgradient methods for mixed binary linear programming; a branch-and-bound approach

1 Portfolio mean and variance

Lecture 5 Principal Minors and the Hessian

THE FUNDAMENTAL THEOREM OF ALGEBRA VIA PROPER MAPS

Transcription:

Approximate Reasoning for Solving Fuzzy Linear Programming Problems Robert Fullér Department of Computer Science, Eötvös Loránd University, P.O.Box 157, H-1502 Budapest 112, Hungary Hans-Jürgen Zimmermann Department of Operations Research, RWTH Aachen Templergraben 64, W-5100 Aachen, Germany Abstract We interpret fuzzy linear programming (FLP) problems (where some or all coefficients can be fuzzy sets and the inequality relations between fuzzy sets can be given by a certain fuzzy relation) as multiple fuzzy reasoning schemes (MFR), where the antecedents of the scheme correspond to the constraints of the FLP problem and the fact of the scheme is the objective of the FLP problem. Then the solution process consists of two steps: first, for every decision variable x R n, we compute the maximizing fuzzy set, MAX(x), via sup-min convolution of the antecedents/constraints and the fact/objective, then an (optimal) solution to FLP problem is any point which produces a maximal element of the set {MAX(x) x R n } (in the sense of the given inequality relation). We show that our solution process for a classical (crisp) LP problem results in a solution in the classical sense, and (under well-chosen inequality relations and objective function) coincides with those suggested by [Buc88, Del87, Ram85, Ver82, Zim76]. Furthermore, we show how to extend the proposed solution principle to non-linear programming problems with fuzzy coefficients. Keywords: Compositional rule of inference, multiple fuzzy reasoning, fuzzy mathematical programming, possibilistic mathematical programming. Published in: Proceedings of 2nd International Workshop on Current Issues in Fuzzy Technologies, Trento, May 28-30, 1992, University of Trento, 1993 45-54. Partially supported by the German Academic Exchange Service (DAAD) and Hungarian Research Foundation OTKA under contracts T 4281, I/3-2152 and T 7598. 1

1 Statement of LP problems with fuzzy coefficients We consider LP problems, in which all of the coefficients are fuzzy quantities (i.e. fuzzy sets of the real line R), of the form maximize c 1 x 1 + + c n x n subject to ã i1 x 1 + +ã in x < (1) n bi,i=1,...,m, where x R n is the vector of decision variables, ã ij, b i and c j are fuzzy quantities, the operations addition and multiplication by a real number of fuzzy quantities are defined by Zadeh s extension principle [Zad75], the inequality relation < is given by a certain fuzzy relation and the objective function is to be maximized in the sense of a given crisp inequality relation between fuzzy quantities. The FLP problem (1) can be stated as follows: Find x R n such that c 1 x 1 + + c n x n c 1 x 1 + + c n x n (2) ã i1 x 1 + +ã in x < n bi,i=1,...,m, i.e. we search for an alternative, x, which maximizes the objective function subject to constraints. We recall now some definitions needed in the following. A fuzzy number ã is a fuzzy quantity with a normalized (i.e. there exists a unique a R, such that µã(a) =1), continuous, fuzzy convex membership function of bounded support. The real number a which maximizes the membership function of the fuzzy number ã called the peak of ã, and we write peak(ã) =a. A symmetric triangular fuzzy number ã denoted by (a, α) is defined as µã(t) = 1 a t /α if a t α and µã(t) =0otherwise, where a R is the peak and 2α >0 is the spread of ã. We shall use the following crisp inequality relations between two fuzzy quantities ã and b ã b iff max{ã, b} = b (3) ã b iff ã b (4) where max is defined by Zadeh s extension principle. The compositional rule of inference scheme with several relations (called Multiple Fuzzy Reasoning Scheme) [Zad73] has the general form Fact X has property P Relation 1: X and Y are in relation W 1...... Relation m: X and Y are in relation W m Consequence: Y has property Q 2

where X and Y are linguistic variables taking their values from fuzzy sets in classical sets U and V, respectively, P and Q are unary fuzzy predicates in U and V, respectively, W i is a binary fuzzy relation in U V, i=1,...,m. The consequence Q is determined by [Zad73] or in detail, m Q = P W i i=1 µ Q (y) = sup min{µ P (x),µ W1 (x, y),...,µ W1 (x, y)}. x U 2 Multiply fuzzy reasoning for solving FLP problems We consider FLP problems as MFR schemes, where the antecedents of the scheme correspond to the constraints of the FLP problem and the fact of the scheme is interpreted as the objective function of the FLP problem. Then the solution process consists of two steps: first, for every decision variable x R n, we compute the maximizing fuzzy set, MAX(x), via sup-min convolution of the antecedents/constraints and the fact/objective, then an (optimal) solution to the FLP problem is any point which produces a maximal element of the set {MAX(x) x R n } (in the sense of the given inequality relation). We interpret the FLP problem maximize subject to c 1 x 1 + + c n x n ã i1 x 1 + +ã in x < n bi,i=1,...,m, (5) as Multiple Fuzzy Reasoning schemes of the form Antecedent 1 Constraint 1 (x) :=ã 11 x 1 + +ã 1n x < n b1...... Antecedent m Constraint m (x) :=ã m1 x 1 + +ã mn x < n bm Fact Goal(x) := c 1 x 1 + + c n x n Consequence MAX(x) where x R n and the consequence (i.e. the maximizing fuzzy set) MAX(x) is computed as follows m MAX(x) =Goal(x) Constraint i (x), i=1 3

i.e. µ MAX(x) (v) = sup min{µ Goal(x) (u),µ Constraint1 (x)(u, v),...,µ Constraintm (u, v)}. u We regard MAX(x) as the (fuzzy) value of the objective function at the point x R n. Then an optimal value of the objective function of problem (5), denoted by M,isdefined as M := sup{max(x) x R n }, (6) where sup is understood in the sense of the given inequality relation. Finally, a solution x R n to problem (5) is obtained by solving the equation MAX(x) =M. The set of solutions of problem (5) is non-empty iff the set of maximizing elements of {MAX(x) x R n } (7) is non-empty. Remark 2.1 To determine the maximum of the set (7) with respect to the inequality relation is usually a very complicated process. However, this problem can lead to a crisp LP problem [Zim78, Buc89], crisp multiple criteria parametric linear programming problem (see e.g. [Del87, Del88, Ver82, Ver84]) or nonlinear mathematical programming problem (see e.g. [Zim86]). If the inequality relation for the objective function is not crisp, then we somehow have to find an element from the set (7) which can be considered as a best choice in the sense of the given fuzzy inequality relation [Ovc89, Orl80, Ram85, Rom89, Rou85, Tan84]. 3 Extension to FMP problems with fuzzy coefficients In this section we show how the proposed approach can be extended to non-linear FMP problems with fuzzy coefficients. Generalizing the classical MP problem maximize g(c, x) subject to f i (a i,x) b i,i=1,...,m, where x R n, c = (c i,...,c k ) and a i = (a i1,...,a il ) are vectors of crisp coefficients, we consider the following FMP problem maximize subject to g( c 1,... c k,x) f i (ã i1,...ã il,x) < b i,i=1,...,m, 4

where x R n, c h, h =1,...,k, ã is, s =1,...,l, and b i are fuzzy quantities, the functions g( c, x) and f i (ã i,x) are defined by Zadeh s extension principle, and the inequality relation < is defined by a certain fuzzy relation. We interpret the above FMP problem as MFR schemes of the form Antecedent 1: Constraint 1 (x) :=f i (ã 11,...ã 1l,x) < b 1...... Antecedent m: Constraint m (x) :=f m (ã m1,...ã ml,x) < b m Fact: Goal(x) :=g( c 1,..., c k,x) Consequence MAX(x) Then the solution process is carried out analogously to the linear case, i.e an optimal value of the objective function, M, isdefined by (6), and a solution x R n is obtained by solving the equation MAX(x) =M. 4 Relation to classical LP problems In this section we show that our solution process for classical LP problems results in a solution in the classical sense. A classical LP problem can be stated as follows max <c,x> (8) Ax b Let X be the set of optimal solutions to (8), and if X is not empty, then let v =<c,x > denote the optimal value of its objective function. In the following ā denotes the characteristic function of the singleton a R, i.e. µā(t) =1if t = a and µā(t) =0otherwise. Consider now the crisp problem (8) with fuzzy singletons and crisp inequality relations max < c, x > (9) Āx b where Ā =[ā i,j], b =( b i ), c =( c j ), and ā i1 x 1 + +ā in x n b i iff a i1 x 1 + + a in x n b i, i.e. 1 if u = v and <a i,x> b i µāi1 x 1 + +ā in x n b i (u, v) = 0 if u = v and <a i,x>> b i 0 if u v (10) 5

The following theorem can be proved directly by using the definition of the inequality relation (10). Theorem 4.1 The sets of solutions of LP problem (8) and FLP problem (9) are equal, and the optimal value of the FLP problem is the characteristic function of the optimal value of the LP problem. 5 Crisp objective and fuzzy constraints FLP problems with crisp inequality relations in fuzzy constraints and crisp objective function can be formulated as follows (see e.g. Negoita s robust programming [Neg81], Ramik and Rimanek s approach [Ram85]) max s.t. <c,x> ã i1 x 1 + +ã in x n b i,i=1,...,m. (11) It is easy to see that problem (11) is equivalent to the crisp MP: where, max x X <c,x> (12) X = {x R n ã i1 x 1 + +ã in x n b i,i=1,...,m}. Now we show that our approach leads to the same crisp MP problem (12). Consider problem (11) with fuzzy singletons in the objective function max < c, x > s.t. ã i1 x 1 + +ã in x n b i,i=1,...,m. where the inequality relation is defined by 1 if u = v and < ã i,x> b i µãi1 x 1 + +ã in x n b i (u, v) = 0 if u = v and < ã i,x>> b i 0 if u v Then we have { 1 if v =< c,x>and x X µ MAX(x) (v) = 0 otherwise Thus, to find a maximizing element of the set {MAX(x) x R n } in the sense of the given inequality relation we have to solve the crisp problem (12). 6

6 Fuzzy objective and crisp constraints Consider the FLP problem with fuzzy coefficients in the objective function and fuzzy singletons in the constraints maximize subject to c 1 x 1 + + c n x n ā i1 x 1 + +ā in x n b i,i=1,...,m, (13) where the inequality relation for constraints is defined by (10) and the objective function is to be maximized in the sense of the inequality relation (3), i.e. < c, x > < c, x > iff max{< c, x >, < c, x >} =< c, x > Then µ MAX(x) (v), v R, is the optimal value of the following crisp MP problem maximize µ c1 x 1 + + c nx n (v) subject to Ax b, and the problem of computing a solution to FLP problem (13) leads to the same crisp multiple objective parametric linear programming problem obtained by Delgado et al. [Del87, Del88] and Verdegay [Ver82, Ver84]. 7 Relation to Buckley s possibilistic LP We show that when the inequality relations in an FLP problem are defined in a possibilistic sense then the optimal value of the objective function is equal to the possibility distribution of the objective function defined by Buckley [Buc88]. Consider a possibilistic LP maximize subject to Z := c 1 x 1 + + c n x n ã i1 x 1 + +ã in x n b i,i=1,...,m. (14) The possibility distribution of the objective function Z, denoted by P oss[z = z], is defined by [Buc88] P oss[z = z] = sup min{p oss[z = z x], P oss[< ã 1,x> b 1 ],...,Poss[< ã m,x> b m ]}, x where P oss[z = z x], the conditional possibility that Z = z given x, isdefined by P oss[z = z x] =µ c1 x 1 + + c nx n (z). The following theorem can be proved directly by using the definitions of P oss[z = z] and µ M (v). 7

Theorem 7.1 For the FLP problem maximize subject to c 1 x 1 + + c n x n ã i1 x 1 + +ã in x < n bi,i=1,...,m. (15) where the inequality relation < is defined by { P oss[< ãi,x> µãi1 x 1 + +ã in x < (u, v) = b i ] if u = v n bi 0 otherwise and the objective function is to be maximized in the sense of the inequality relation (4), the following equality holds µ M (v) =P oss[z = v], v R n. So, if the inequality relations for constraints are defined in a possibilistic sense and the objective function is to be maximized in relation (4) then the optimal value of the objective function of problem (15) is equal to the possibility distribution of the objective function of possibilistic LP (14). In a similar manner it can be swhon that the proposed solution concept is a generalization of Zimmermann s method [Zim76] for solving LP problems with soft constraints. 8 Example We illustrate our approach by the following FLP problem maximize cx subject to ãx < b, x 0, (16) where ã = (1, 1), b = (2, 1) and c = (3, 1) are fuzzy numbers of symmetric triangular form, the inequality relation < is defined in a possibilistic sense, i.e. { P oss[ãx b] if u = v, µãx < b(u, v) = 0 otherwise and the inequality relation for the values of the objective function is defined by cx cx iff peak( cx) peak( cx ). It is easy to compute that 4 v/2 if v/x [3, 4] µ M (v) =µ MAX(2) (v) = v/x 2 if v/x [2, 3] 0 otherwise 8

so, the unique solution to (16) is x =2. This result can be interpreted as: the solution to problem (16) is equal to the solution to the following crisp LP problem max 3x s.t. x 2, x 0, where the coefficients are the peaks of the fuzzy coefficients of problem (16), and the optimal value of problem (16), which can be called v is approximately equal to 6, can be considered as an approximation of the optimal value of the crisp problem v =6. 1 4 6 8 Figure 1. The membership function of the optimal value of problem (16). Concluding remarks. We have interpreted FLP problems with fuzzy coefficients as MFR schemes, where the antecedents of the scheme correspond to the constraints of the FLP problem and the fact of the scheme is the objective of the FLP problem. Using the compositional rule of inference, we have shown a method for finding an optimal value of the objective function and an optimal solution. In the general case the computerized implementation of the proposed solution principle is not easy. To compute MAX(x) we have to solve a generally non-convex and non-differentiable mathematical programming problem. However, the stability property of the consequence in MFR schemes under small changes of the membership function of the antecedents [Ful91] guarantees that small rounding errors of digital computation and small errors of measurement in membership functions of the coefficients of the FLP problem can cause only a small deviation in the membership function of the consequence, MAX(x), i.e. every successive approximation method can be applied to the computation of the linguistic approximation of the exact MAX(x). As was pointed out in Section 2, to find an optimal value of the objective function, M, from the equation (6) can be a very complicated process (for related works see e.g. [Car82, Car87, Neg81, Orl80, Ver82, Ver84, Zim85, Zim86) and very often we have to put up with a compromise solution [Rom89]. An efficient fuzzy-reasoning-based method is needed for the exact computation of M. Our solution principle can be applied to multiple criteria mathematical pro- 9

gramming problems with fuzzy coefficients. This topic will be the subject of future research. Acknowledgements The authors are indebted to Brigitte Werners of RWTH Aachen and the participants of the Workshop on Current Issues on Fuzzy Technologies 1992 for their valuable comments and helpful suggestions. References [Buc88] [Buc89] [Car82] [Car87] [Del87] [Del88] [Dub80] [Ful91] [Neg81] [Orl80] J.J.Buckley, Possibilistic linear programming with triangular fuzzy numbers, Fuzzy sets and Systems, 26(1988) 135-138. J.J.Buckley, Solving possibilistic linear programming problems, Fuzzy Sets and Systems, 31(1989) 329-341. C.Carlsson, Tackling an MCDM-problem with the help of some results from fuzzy sets theory, European Journal of Operational research, 3(1982) 270-281. C.Carlsson, Approximate reasoning for solving fuzzy MCDM problems, Cybernetics and Systems: An International Journal, 18(1987) 35-48. M.Delgado,J.L.Verdegay and M.A.Vila, Imprecise costs in mathematical programming problems, Control and Cybernetics, 16(1987) 114-121. M.Delgado,J.L.Verdegay and M.A.Vila, A procedure for ranking fuzzy numbers using fuzzy relations, Fuzzy Sets and Systems, 26(1988) 49-62. D.Dubois and H.Prade, Systems of linear fuzzy constraints, Fuzzy Sets and Systems 3(1980) 37-48. R.Fullér and H.-J.Zimmermann, On Zadeh s compositional rule of inference, in: R.Lowen and M.Roubens eds., Proc. of Fourth IFSA Congress, Vol. Artifical Intelligence, Brussels, 1991 41-44. C.V.Negoita, The current interest in fuzzy optimization, Fuzzy Sets and System, 6(1981) 261-269. S.A.Orlovski, On formalization of a general fuzzy mathematical problem, Fuzzy Sets and Systems, 3(1980) 311-321. 10

[Ovc89] [Ram85] S.V.Ovchinnikov, Transitive fuzzy orderings of fuzzy numbers, Fuzzy Sets and Systems, 30(1989) 283-295. J.Ramik and J.Rimanek, Inequality relation between fuzzy numbers and its use in fuzzy optimization, Fuzzy Sets and Systems, 16(1985) 123-138. [Rom89] H.Rommelfanger, Entscheiden bei Unschärfe, Springer-Verlag, Berlin, 1989. [Tan84] [Ver82] [Ver84] [Zad73] [Zad75] [Zim78] [Zim85] [Zim86] [Zim76] H.Tanaka and K.Asai, Fuzzy solution in fuzzy linear programming problems, IEEE Transactions on Systems, Man, and Cybernetics, Vol. SMC-14, 1984 325-328. J.L.Verdegay, Fuzzy mathematical programming, in: M.M.Gupta and E.Sanchez eds., Fuzzy Information and Decision Processes, North- Holland, 1982. J.L.Verdegay, Applications of fuzzy optimization in operational research, Control and Cybernetics, 13(1984) 230-239. L.A.Zadeh, Outline of a new approach to the analysis of complex systems and decision processes, IEEE Trans. on Systems, Man and Cybernetics, Vol.SMC-3, No.1, 1973 28-44. 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. H.-J.Zimmermann, Fuzzy programming and linear programming with several objective functions, Fuzzy Sets and Systems,1(1978) 45-55. H.-J.Zimmermann, Applications of fuzzy set theory to mathematical programming, Information Sciences 36(1985) 29-58. H.-J.Zimmermann, Fuzzy Sets Theory and Mathematical Programming, in: A.Jones et al. (eds.), Fuzzy Sets Theory and Applications, D.Reidel Publishing Company, 1986 99-114. H.-J.Zimmermann, Description and optimization of fuzzy system, International Journal of General Systems, 2(1976) 209-215. 11