Int. J. Contemp. Math. Sciences, Vol. 7, 2012, no. 19, 897 - 908

A Study on Fuzzy Complex Linear

Programming Problems

Youness, E. A.(1) and Mekawy, I. M.*(2)

(1) Department of Mathematics, Faculty of Science Tanta University, Tanta, Egypt [email protected]

(2) Ministry of Education A. R. E., Al-Imam Muhammad Ibn Sauid Islamic University, Scientific Institute in Jouf Jouf, Saudi Arabia

Abstract. In this paper we consider a certain kind of linear programming problems. This kind, in which fuzzy complex is the coefficient of the objective function, is called fuzzy complex linear programming problem. Solutions of this kind of problems will be characterized by deriving its fuzzy Kuhn-Tucker conditions. The results include a duality relation between α -cut programming and its dual are discussed. Also, we introduce the proof of Kuhn-Tucker stationary- point necessary optimality theorem for our problem.

Keywords: Cone, polyhedral cone, , fuzzy number, α -cut programming fuzzy , complex programming.

1. Introduction

Since the concept of fuzzy complex was first introduced by J. J. Buckley in 1989 [6], many papers were devoted to studying the problems of the concept of fuzzy complex numbers. This new branch subject will be widely applied in fuzzy system theory, especially in fuzzy mathematical programming, and will also be widely applied in complex mathematical programming. It is well known that, complex programming, the extension to complex variables and functions of mathematical programming was initiated by N. Levinson in [9], where a duality theory for complex linear programming is given and the basic theorem of linear inequalities are extended to the complex case. Similar results were previously developed by Bellman and Fan in [1], for Hermitian matrices. This work of Levinson was continued by Hanson and Mond in [3], where it is

* Corresponding Author, e-mail: [email protected] 898 Youness, E. A. and Mekawy, I. M.

considered as an extensions to quadratic and non linear programming. The fuzzy set theory has been applied to many disciplines such as control theory and management sciences, mathematical modeling, operations research and many industrial applications. The concept of fuzzy mathematical programming on general level was first proposed by Tanaka et al. [4]. In the framework of the fuzzy decision of Bellman and Zadeh [12]. Zimmermann [5] first introduced fuzzy linear programming as conventional linear programming. He considered problem with a fuzzy goal and fuzzy constraints, used linear membership functions and the min operator as an aggregator for these functions, and assigned an equivalent problem to fuzzy linear programming.This study focuses on fuzzy complex linear programming (FCLP) problems. Hence, first some important concepts on fuzzy and complex mathematical programming are mentioned. This paper organized in 5 sections. In section 2, we give some necessary notations and definitions of fuzzy set theory, and complex mathematical programming problems. The concept of fuzzy complex numbers is introduced in section (3). Formulation of fuzzy complex linear programming, and deriving Kuhn-Tucker stationary point problem are presented in section 4. Finally, section 5 for conclusion.

2. Notations and Definitions

2.1. Notations: - CRnn() the n-dimensional complex (real) vector space over the field R n . - CRmn××[] mn the mn× complex [real] matrices. mn× - For Aa=∈()ij C :

Aa≡−()ij the conjugate, T Aa≡−()ji the transpose, AAHT≡−the conjugate transpose. nn - For x =∈()xCyCi , ∈: (,x yyx )≡−H the inner product of x and y, n ||(||)xxR≡∈−i the absolute value,

xx≡−()i the conjugate, n RexxR≡∈− ( Rei ) the real part, of x n ImxxR≡∈− ( Imi ) the imaginary part, of x

argxx≡− ( argi ) the argument of x. - For a nonempty set SC⊂ n : SyCxS* ≡∈{:n ∈⇒ Re(,)0} yx ≥−the dual (also polar) of S nn n - R+ ≡∈{:0)(1,...,)}xR xi ≥ i = n the nonnegative orthant of R n n x ≥ y denoted x −∈yR+ , for (,x yR )∈ . Fuzzy complex linear programming problems 899

- For an analytic function f : CCn → and a point zC0 ∈ n

00⎛⎞∂ f 0 ∇≡z f ()zzin⎜⎟ (),(1,...,) = the gradient of f at z . ⎝⎠∂ z i 2.2. Definitions Definition 2.2.1. [1]. A non empty set SC⊂ n is: (a) convex if 01≤≤⇒λ λλSSS +− (1) ⊂, (b) a cone if 0 ≤ λ ⇒⊂λ SS, (c) a polyhedral cone if for some positive k and AC∈ nk× : kk SAR==++{: AxxR ∈ }, i.e. S is generated by finitely many vectors (the columns of A). Definition 2.2.2. [7]. A fuzzy set a% in R is a set of ordered pairs:

ax% =∈{ (,μa% ( xxR )): }} ,

μa% ()x is called the membership function of x in a%

μa% ():xR→ [0,1], if μa% ()1x = , the fuzzy set a% is called normal. Definition 2.2.3. [7]. The support of a fuzzy set a% on R is the crisp set of all x ∈ R such that μa% ()0x > . Definition 2.2.4. The α - level set (cut set) of a fuzzy set a% is defined as an ordinary set aα for which the degree of its membership function exceeds the level α :

axα =≥∈{:μa% () xαα }, [0,1]. Definition 2.2.5. [4]. A fuzzy set a% on R is convex if

μaaa%%%((1))min{(),()},,λλx +−yxyxyR ≥ μμ ∈, and λ ∈[0,1]. Note that, a fuzzy set is convex if all α -cuts are convex. Definition 2.2.6. [5]. A fuzzy number a% is a convex normalized fuzzy set on the real line R such that: 0 (1) It exists at least one x 0 ∈ R which μa% ()1x = ,

(2) μa% ()x is piecewise continuous. From the definition of a fuzzy number a% , it is significant to note that, the α -level set aα of a fuzzy number a% can be represented by the closed interval which depends on the value of α . Namely, −+ axRxα =∈{ :μa% () ≥=ααα }[ a (),()] a , where a− ()α or a+ ()α represents the left or right extreme point of the α -level set aα , respectively

3. Fuzzy Complex Numbers [2, 6, 8, 13]

The concept of fuzzy complex numbers was first introduced by J. J. Buckley in 1989, many papers were devoted to studying the problems of the concept of fuzzy 900 Youness, E. A. and Mekawy, I. M.

complex numbers. Uncertainty of complex valued physical quantities caib=+ can be described by complex fuzzy sets, such sets can be described by membership functions μ (,ab ) which map the universe of discourse (complex plane) into the interval [0,1]. The problem with this description is that it is difficult to directly translate into words from natural language. To make this translation easier, several authors have proposed to use, instead of a single membership function for describing the complex number, several membership function which describe different real valued characteristics of this number, such as its real part, its imaginary part, its absolute value, etc. Thus, a natural idea is to represent a complex fuzzy number by describing two real fuzzy numbers: a% and b% characterized by the corresponding membership functions μa% (),abμb% (). In this approach, for every complex value aib+ , i.e., for every pair (,ab ), the degree μ (,ab ) with which this complex value is possible can be defined as

μcc%%()cabab==μμμ (, ) min( a % (),b% ()).

3.1 Preliminary Concepts

Let R be a set of real numbers, CxiyxRyRi= {:,,1}+∈∈=− a field of complex numbers, a finite closed interval XXX= [,− + ] is called a closed interval number on R, I ()R denotes the set of all closed interval numbers on R. For arbitrary intervals XXX=[,− + ], YYY=[,−+ ]∈ R, Z = XiY+={x +∈iy C :,xXyYi∈∈=− , 1} is called a closed complex interval number, IC()= {:,(),1}ZXiYXYIRi=+ ∈ =− denotes the set of all closed complex interval numbers on C, ∗=′′{, +−×÷ ,, } is a binary operation on I ()C , (when

∗=Z12 ÷ZZ, 0 ∉ supp 2 ) . −+ −+ Definition 3.1. For ∀=+=ZXiYXXkk k[, kk ][, + iYY kk ](),(1,2) ∈ ICk = , − + −+ zz12∗={: z ∃ (, zz 12 ) ∈× ZZzzz 12 , =∗ 12 } and XXX= [, ], YYY=[, ] ∈ I ()R . Specifically, for kR∈=+∞+ [0, ), let kX⋅=[, kXkX− + ], let kZ⋅ =+()() kXikY = −+ −+ [,kX kX ][,]+ i kY kY , let Z kk= XiY+ k (1,2)(),kICZZ= ∈⊆⇔12

XXYY1212⊆⊆, . If XY, are convex sets on R, Z ⊆ C and Z =+∈{:,}xiyCxX ∈ yY ∈, then Z is a convex set on C. Definition 3.2. Let C be a field of complex number, mapping ZC% :[0,1]→ is called a fuzzy complex set, Z% ()z is called the membership function of fuzzy set Z% for zFC,(){:=→ Z%% Z : C [0,1]} denotes all fuzzy complex sets on C.

Definition 3.3. Z%%%α ==+∈{:()()} z x iyC Zz = Zx + iy ≥α is called α -cut set of Z% .

Fuzzy complex linear programming problems 901

Definition 3.4. ZZzxiyCZzZxiy%%0 ===+∈=+>supp { : %% ( ) ( ) 0} is called support set of Z% . Definition 3.5. Z% ∈ FC() is called a convex fuzzy complex set on C, if and only if for ∀∈α [0,1], Z%α is a convex complex set on C. Definition 3.6. Z% ∈ FC() is a normal fuzzy complex set on C, if and only if {:()()1}zCZz∈=+=≠%% Zxiy φ . Definition 3.7. A normal convex fuzzy complex set on C is called a fuzzy complex number.

3.2. Buckley's Membership Function Description.

In order to describe a complex number Caib= + , we must describe two real numbers: its real part a and its imaginary pat b. Thus, a natural idea is to represent a complex fuzzy number caib%%=+% , by describing two real fuzzy numbers: a% and b% (see [2]) characterized by the corresponding membership functions μa% ()a and μb% ()b . In this approach, for every complex value aib+ , i.e., for every pair (,ab ), the degree μcc%%()cab= μ (, ) with which this complex value is possible can be defined as

μcc%%()cabab==μμμ (, ) min((), a %b% ()). −+ Then , for each α ∈[0,1], the α -cut for the real part a% is an interval [(),()]aaα α , the α -cut for the imaginary part is also an interval [(),()]bb−+α α , and hence, the α -cut for the resulting 2-D membership function

μca%%()cab= min((),μμb% ()), is a rectangular "box" [aa−+ (),()][α ααα× bb −+ (),()]. The boundary of this box consists of two straight line segments, which are parallel to the a-axis, and of two straight line segments which are parallel to the b-axis.

4. Fuzzy Complex Linear Programming

Consider the complex linear programming problem [1, 3, 9, 10] ⎧ min Re (cx , ) ⎪ ()s.tP1 ⎨ ⎪ ⎩ Ax−∈ b T,, x ∈ S where, A ∈∈CcCmn× , n, and let SC⊂ n and TC⊂ m be closed convex cones. 902 Youness, E. A. and Mekawy, I. M.

The fuzzy complex linear programming can be formulated as: ⎧ min Re (cx% , ) ⎪ ()s.tP% ⎨ ⎪ ⎩ Ax−∈ b T,, x ∈ S % where caib%%=+ is a fuzzy complex number, μca%%()cab= min((),μμb% ()), for each −+ α ∈[0,1], the α -cut for the real part a% is an interval [(),()]aaα α , and the α -cut for b% is [(),()]bb−+α α , and hence the α -cut for the resulting 2-D membership function μcc%%()cabab==μμμ (, ) min((), a %b% ()) is [aa−+ (),()][α ααα× bb −+ (),()], and it can be written in the following form

Lcα =≥{:μc% () cα }, where caib=+ , μca%%()cab= min((),μμb% ()), α = (,αα12 ), μa% ()a ≥ α1 and μb% ()b ≥ α2 . The fuzzy complex linear programming ()P% can be converted to the deterministic α - cut programming as ⎧ min Re (cx , ) ⎪ ⎪ s. t ()Pα ⎨ ⎪ Ax−∈ b T,, x ∈ S ⎪ ⎩ cL∈ α .

Choose cLˆ ∈ α and characterize the solution of ()Pα corresponding to cˆ , i.e. we will solve the problem ⎧ min Re (cxˆ , ) ⎪ ()s.tP ⎨ ⎪ ⎩ Ax−∈ b T,, x ∈ S The dual problem of problem ()P is ⎧ max Re (by , ) ⎪ ()s.tD ⎨ ⎪ H ** ⎩ cAˆ −∈∈ y S, yT where S * and T * are dual cones of S and T respectively. Definition 4.1. A vector x 0 ∈C n is (i) a feasible solution of ()P is A xbTxS00−∈, ∈ , (ii) an optimal solution of ()P if x 0 is feasible and Re (cxˆ ,0 ) = min { Re (cxˆ , ); xis feasible }. Definition 4.2. The problem ()P is (i) consistent if it has feasible solutions,

(ii) unbounded if it is consistent, and if it has feasible solutions {;xkk = 1,2,...}

with Re (cxˆ ,k ) →−∞.

Fuzzy complex linear programming problems 903

Consistency and boundedness of ()D and feasibility and optimality of its solutions, are similarly defined. Definition 4.3. The Lagrangian of the problems ()P and ()D is L (,xy )=−− Re{(, cxˆ )(, yAxb )} =+−Re{(by , ) ( cˆ AH yx , )} Definition 4.4. The point (,x 00yST )∈ × * is a saddle point of L (,xy ) with respect to ST× * if L (,xyLxy0000 )≤≤ (, ) Lxy (, ) for all x ∈∈SyT, * . A duality relations between ()P and ()D and a characterization of there optimal solutions, (if such solutions exist) are given in the following theorem. Theorem 4.1. [6]. Let S and T in problems ()P and ()D be polyhedral cones, then: (a) If one of the problems is inconsistent then, the other is inconsistent or unbounded. (b) Let the two problems be consistent, and let x 0 be a feasible solution of ()P and y 0 be a feasible of ()D . Then Re (cxˆ ,00 )≥ Re ( by , ) . (c) If both ()P and ()D are consistent, then they have optimal solutions and their optimal values are equal. (d) Let x 0 and y 0 be feasible solutions of ()P and ()D respectively. Then x 0 and y 0 are optimal if and only if Re{(Ax00− b , y )+− ( cˆ AH y 00 , x )} = 0 , or equivalently if and only if Re (Ax00−=−= b , y ) ( cˆ AH y 00 , x ) 0 . (e) The vectors x 0 ∈C n and yC0 ∈ m are optimal solutions of ()P and ()D respectively if and only if the point (,x 00y ) is a saddle point of L (,xy ) with respect to ST× * , in which case L (,xy00 )Re(,)Re(,)== cxˆ 0 by 0. 0 Now, let us return to the problem ()Pα , suppose that x is a solution, the problem becomes: ⎧ min Re (cx ,0 ) ⎪ ⎨ s. t ⎪ 00 ⎩ Ax−∈ b T,,, x ∈ S c ∈ Lα which is equivalent to ⎧ min Re (cx ,0 ) ⎪ ⎨ s. t ⎪ cL∈ , ⎩ α % where Lcα =≥={:μc% () cαααα },(,12 ), caib%%=+ , μca%%()cab= min{(),μμb% ()}. The problem becomes 904 Youness, E. A. and Mekawy, I. M.

⎧ min Re (cx ,0 ) ⎪ ⎨ s. t ⎪ αμ−≤()c 0, ⎩ c% where μca%%()cab= min((),μμb% ()), α = (,αα12 ), μa% ()a ≥ α1 , μb% ()b ≥ α2 . We can define the following problems: 1- The minimization problem ( MP ). ⎧ FindcL* ∈ , if it exists, such that ⎪ α MP ⎨ Re (cx*0 , )=∈=−≤ min Re ( cx , 0 ), c * L { c :αμ ( c ) 0}. ⎪ α c% ⎩ cL∈ α 2- The local minimization problem ()L MP ⎧ Find ancL** in , if it exists, such that for some open ball Bc ( ) ⎪ αδ ⎪ * LMP⎨ around c with radius δ >0 ⎪ cB∈⇒≤() c**00 L Re(, c x )Re(,) cx ⎩⎪ δαI **m 3- The Kuhn-Tucker stationary-point problem ()KTP : Find cLuR∈∈α , , if they exist, such that ⎧ψαμ(,cu**0* )Re(,=+− cx ) u u ( c ), ⎪ c% ** *0 * * ⎪ ∇=⇒∇−∇=ccccψμ(,cu )0 Re(, cx ) u% ( c )0, ⎪ ** * KTP⎨ ∇≤⇒−≤ucψαμ(, c u )0% ( c )0, ⎪ **** * ⎪ ucuu∇=⇒−=ucψαμ(, )0 [% ( c )]0, ⎪ u * ≥ 0 ⎩⎪ In the following, we introduce some constraint qualifications, which, we need it in our study, see [8] 1- Slater's Constraint Qualification. n Let Lα be a convex set in C , and μc% ()c be concave on Lα .

α −≤μc% ()c 0 is said to satisfy slater's constraint qualification if there exists an * * cL∈ α such that αμ−

α −≤μc% ()c 0 is said to satisfy Karlin's constraint qualification if there exists no

PRP∈≥,0 such that PPα −≥μc% () c 0 for all cL∈ α . 3- The Strict Constraint Qualification. n Let Lα be a convex set in C , and μc% ()c is concave on Lα .

α −≤μc% ()c 0 is said to satisfy strict constraint qualification if α − μc% ()c is strictly 12 convex at distinct points cc, ∈ Lα . 4- The Kuhn-Tucker Constraint Qualification. n Let Lα be an open set in C , and μc% ()c defined on Lα , α − μc% ()c ≤ 0 is said to * satisfy the Kuhn-Tucker constraint qualification at cL∈ α if α − μc% ()c is Fuzzy complex linear programming problems 905

* n differentiable at c and if yC∈ ,()Re0∇⋅≤ccμ % c y implies there exists an n- dimensional vector function e defined on the interval [0, 1] such that (a) ec(0)= * ,

(b) eL()τ ∈ α for 01≤≤τ , d (c) e is differentiable at τ = 0 and ((0))ey= λ for some λ > 0 . d τ 5- The Arrow-Hurwicz- Uzawa Constraint Qualification. n Let Lα be an open set in C , and μc% ()c defined on Lα , α − μc% ()c is said to satisfy * the Arrow-Hurwicz-Uzawa constraint qualification at cL∈ α if α − μc% ()c is * n differentiable at c and if ∇ccμ % ()Recz has a solution zC∈ . 6- The Reverse Convex Constraint Qualification. n Let Lα be an open set in C , and α − μc% ()c defined on Lα , α − μc% ()c is said to * satisfy the Arrow-Hurwicz-Uzawa constraint qualification at cL∈ α if * * α − μc% ()c is differentiable at c , and either α − μc% ()c is concave at c or n α − μc% ()c is linear in R . n 0 Theorem 4.2. Let Lα be an open subset of C , let Re (cx , ) be convex and μc% ()c * be concave on Lα , cL∈ α solve the problem min Re (cx ,0 ) s. t.

cL∈=α {: cα −μc% ()0} c ≤, 0 * and let Re (cx , ) and α − μc% ()c be differentiable at c and let α − μc% ()c satisfies one of the constraint qualification 1-6. Then, there exists a uR* ∈ such that (,cu** ) satisfies: *0 * * Re∇−∇=ccc (cx , ) uμ % ( c ) 0 , and ** uc[()]0αμ−=c% . * ˆ * nT Proof: Let c solve LMP with δ = δ . If αμ− c% ()0c < , then, for yC∈ ,1 yy= , we have: **** αμ−+=−−∇+ccc%%%()cy δ αμ ()[()(,)] c δ μ cycy β δ ** ** ⇒+−=∇+μδμcccc%%()()[()(,)]cy c δμ % cycy βδ * * Since αμ−

Since y is an arbitrary vector in C n satisfying yyT = 1 we conclude from this last inequality, by taking ye=± i , where eCin∈ is a vector with one in the ith position and zero elsewhere, that *0 Re∇=c (cx , ) 0 . Hence c * and u * = 0 satisfy that *0 * *0 Re∇−∇=cc (cx , ) u ( cx , ) 0 , and ** uc[()]0αμ−=c% . * If αμ−=c% ()0c . Let α − μc% ()c satisfy the Kuhn-Tucker constraint qualification at c * , and let yC∈ n satisfy * −∇c ()Re0cy ≤ , there exists an n-dimensional vector function e defined on [0, 1] such that * eceL(0)=∈ , (τ ) α for 01≤≤τ , de((0)) e is differentiable at τ = 0 , and = λ Re y for some λ > 0 . Hence for 01≤≤τ dt

⎡⎤dei (0) eeii()ττ=+ (0)⎢⎥ + γτ i (0,) for in= 1, ..., . ⎣⎦d τ where limγ i ( 0,τ )= 0 . Hence by taking τ small enough, say 01< τ <<τˆ , we have τ →0 * * that eBc()τ ∈ δ ( ) since eL()τ ∈ α , for 01≤ τ ≤ and c solves LMP, we have that Re (ex (τ ),00 )≥ Re ( ex ( 0 ), ) for 0 < τ < τˆ . Hence by the chain rule and the differentiability of Re (cx ,0 ) at c * , we have for 0 <<τ τˆ that de(0 ) 0Re((),)Re((0),)≤−exτ 00 ex=∇Re (ex ( 0 ),0 )τ +βττ ( 0, ) , d τ where limβ ( 0,τ )= 0 . Hence τ →0 de(0) Re∇+≥ (ex ( 0 ),0 )βτ ( 0, ) 0 for 0 < τ < τˆ . d τ Taking the lim as τ approaches zero gives de(0) Re∇≥ (ex ( 0 ),0 ) 0 . d τ de(0) Since ec(0)= * and = λ Re y for some λ > 0 , we have that: d τ *0 Re (∇≥c (cxy , ) ) 0 . Hence, w have shown that if **0 Re (−∇ccμ % (cy ) ) ≤ 0 ⇒ Re ( ∇ c ( c , x ) y ) ≥ 0 , or that Re (∇< (cxy*0 , ) ) 0 c has no solution yC∈ n . * Re (−∇ccμ % (cy ) ) ≤ 0 Fuzzy complex linear programming problems 907

* * Hence by Motzkin's theorem of alternative in [7], there exists an r0 and r such that r * rcxrc**0**Re∇−∇= ( , )μ ( ) 0 , or Re∇ (cx*0 , )−∇μ ( c * ) = 0 , 0 ccc% ccc* % r0 ** * * * rr0 ≥≥0, 0 . Since r0 is a , r0 ≥ 0 means r0 > 0 , then by defining r * u * = , we have that * r0 *0 * * Re∇−∇=ccc (cx , ) uμ % ( c ) 0 , and ** uc(())0αμ−=c% .

5. Conclusions

In this work, we can find a certain kind of mathematical programming problems called fuzzy complex linear programming by using the concept of fuzzy complex number in complex mathematical programming and α -cut of a fuzzy complex number. Also, we introduce the proof of Kuhn-Tucker stationary-point necessary optimality theorem for our problem.

References

[1] A . Berman, Lecture notes in economics and mathematical systems, Springer Verlag, Berlin, Heidelberg, New York, 1973. [ 2 ] A. Kaufmann and M. M Gupta , Introduction to fuzzy arithmetic, Van Nostrand Reinhold Co., N. Y., 1985. [3] B.Mond and M. A Hanson, On duality for real and complex programming problems, J. Math. Anal. Appl., 24, (1968), 307-312. [4] H.Tanaka, T. Dkuda and K. Asai, On fuzzy mathematical programming. The journal of Cybernetics 3, (1974), 37-46. [ 5] H. J. Zimmermann , Fuzzy programming and linear programming with several objective functions", Fuzzy Sets and System 1 (1978) , 45-55. [6] J. J Buckley, Fuzzy complex numbers. Fuzzy Sets and Systems, 33 (1989), 333- 345. [7] M. Sakawa, Fuzzy sets and interactive multiobjective optimization, plenum press, New York and London, 1973. [8] M. Shengquan and C. Chun, Fuzzy complex analysis, the Ethnic House, Beijing 2001. [ 9 ] N Levinson , Linear programming in complex space, J. Math. Appl., 14, (1966), 44-62. [10] O. L Mangasarian, Non linear programming, McGraw Hill, New York (1969).

908 Youness, E. A. and Mekawy, I. M.

[11] R. E. Bellman and K . Fan , On system of linear inequalities in Hermitian matrix variables, proceedings of symposia in pure mathematics, volume VII, Convexity (edited by V. Klee), Amer, Math. Soc., Providence, Rhode Island, (1963), 1-11. [12] R. E. Bellman. and L. A. Zadeh, Decision making in a fuzzy environment, Management Sci. 17 (1970), 141-164. [13] S. C. Fang and C. F Hu, Linear programming with fuzzy coefficients in constraint, Comput. Math. Appl. 37 (1999), 63-76.

Received: October, 2011