sets and systems

ELSEVIER Fuzzy Sets and Systems 99 (1998) 311-324

Fuzzy random variable-valued exponential function,

logarithmic function and power function
Qiao Zhong"'*, Zhang Yue b, Wang Guangyuan c
"Department of Management Engineering, China Agricultural University (Eastern Campus), P.O. Box 108, No. 17 Qinghua Donglu,
Beijing 100083, China
b Tsinghua University, Beijing 100084, China
cHarbin University of Architecture and Engineering, Harbin 150006, Heilongfiang, China
Received December 1995; revised November 1996

Abstract

This paper introduces concepts of fuzzy random variable-valued exponential function, logarithmic function and power
function. Their fundamental properties are also discussed. © 1998 Elsevier Science B.V. All rights reserved

Keywords: Fuzzy random variable; Fuzzy random variable-valued function; Fuzzy number; Analysis

1. Introduction

The concept of a fuzzy random variable was first introduced by Kwakernaak [2]. Puri and Ralescu [4]
proposed another notion of a fuzzy random variable. Wang and Zhang [9, 12] presented a concept of a fuzzy
random variable slightly different than those presented in [2, 4]. It can be proven that these concepts are
equivalent under some conditions. Qiao and Wang [5-11] further studied some fundamental properties of
fuzzy random variables. In order to study the theory of fuzzy random programming which is used to make
decisions in a fuzzy random system, we shall introduce concepts of fuzzy random variable-valued exponential
function, logarithmic function and power function, and discuss their fundamental properties in this paper.
The following concepts and results are introduced from [5-12].
Let ~-o(R) denote the set of all bounded closed fuzzy numbers, i.e. if f ~ ~o(R), then f satisfies
(1) {xlx~R, ~(x)= 1} # 0 ; (1.1)
(2) for every ~ E (0, 1],

~= = { x l x ~ R, F(x) >~ ~} = [r;, r+~] (1.2)

is a finite closed interval on R, where R = ( - ~, ~).

* Corresponding author.

0165-0114/98/$19.00 © 1998 Elsevier Science B.V. All rights reserved

PII S 0 1 6 5 - 0 1 1 4 ( 9 6 ) 0 0 3 9 7 - 1
312 Qiao Zhong et al. / Fuzzy Sets and Systems 99 (1998) 311-324

Definition 1.1. A mapping 4: O ~ ~o(R) is called a fuzzy random variable on the probability space (f2, M, P),
if for every B e ~, we have

{wl w ~ t2, 4,(w)c~B ~ 0} e M for every a e (0, 1], (1.3)

where ~ is the set of Borel subsets on R, 4,(w) = { x l x e R, 4(w) (x)/> ~} is the level set of 4(w). Denote the set
of all fuzzy random variables on (O, ~1, P) by FR(Y2).

Theorem 1.2. ~i e FR(t2) if and only if ?t~(w) = [a; (w), a~+(w)] is a closed random interval number on (f2, ~ , P)
for every ~ ~ (0, 1] (namely, a; (w) and a+~(w) are two random variables such that a; (w) <~a +(w) everywhere
on I2).

Definition 1.3. Let 4, ~; e FR(~2). We define that

(1) 4 ~< ~; if and only if 4~(w) ~< ~;~(w), i.e. a~- (w) ~< b~- (w) and a~+ (w) ~< b~+ (w) for every w e t2, for any
E(O, 1].
(2) 4 < ~; if and only if 4~(w) < ~;~(w), i.e. a~- (w) < b~- (w) and a~+ (w) < b~+ (w) for every w ~ f2, for any
~ ( 0 , 1].

Definition 1.4. Let • be an algebraic operation on ~-0 (R) (* may b e " + " , " - ", ". ", "/", etc; see [5-12]). The
algebraic operation • on FR(f2) may be defined by

(4 • ~)(w)~S(w) • ~(w) for any w e O, (1.4)

where 4, ~; e FR(f2).

For +, -, , / of closed interval numbers, readers may see [3-10]. Evidently,

R ~ ~-o(R) c FR(I2), and R c R(t2) ~ FR(~2), where R(O) is the set of all random variables on (f2, •, P).

2. Fuzzy random variable-valued exponential function

In this section, we shall introduce the concept of fuzzy random variable-valued exponential function, and
discuss its fundamental properties. Suppose ~ is any nonempty subset of FR(I2).

Definition 2.1. The mapping f: ~ ~ FR(~2) (~--~f(~)) is called a fuzzy random variable-valued function
defined on ~, where ~ is said to be the domain of definition forf(~), and ~ is called the independent variable.

It is clear that f(~) is also a fuzzy random variable in FR(I2) for every ~ ~ ~.

Example 2.2. Let 4 ~ FR(I2) be a given fuzzy random variable; then

fl(~) = 4 . ~ for any ~ ~ ~, (2.1)

f2(x) ----4 -b X for any Y ~ ~ (2.2)

are two fuzzy random variable-valued functions defined on ~,

It is clear that we may structure maps from ~ to FR(t2) by all kinds of methods. In the following, we shall
use the extension principle to define the fuzzy random variable-valued exponential function.
 Qiao Zhong et al. / Fuzzy Sets and Systems 99 (1998) 311-324 313

2.1. Fuzzy random variable-valued exponential function exp

L e m m a 2.3. Suppose that f (x) (x E R) is an ordinary real-valued continuous function, 5 • ~o (R). If we define

f(5)g U ef(5~), (2.3)

a~(0, 11

then
(1) for any a e (0, 1],

(2) f(5) • ,~o(R).

Proof. For any u • (0, 1], since

f(5=) --- {f(x) l x • 5= -- [a~-, a~+]},

and f ( x ) is an ordinary continuous function, then there exist A x~a,f(x) and Vx+a,f ( x ) such that

Furthermore, let 5(Xo) = 1 and Yo =f(xo), where Xo • R; then

Yo •f(51) = {f(x)Lx • 51 = [al, a;-]},
it follows that

f(a)(yo) = V a'f(a,)(yo)= V c¢= 1.

~¢(0, 1] yo ~ f (5,)

That is to say, f(5) is also a bounded closed fuzzy number in ~-o(R). []

Definition 2.4. Let exp x = e ~ (where x • R = (-0% oo)) be the ordinary exponential function. For any
• FR(Y2), we define

exp.& U ~exp~, (2.5)

ae(O, 11

where exp ~ = {exp x Ix • ~ = [x~-, x~+]}, a s (0, 1]. exp ~ is called the fuzzy random variable-valued
exponential function to the base e.

Theorem 2.5. exp ~ has the following properties:

(1) (exp ~)(w) • .~o(R) for every w ~ f2.
(2) (exp ~)~ = exp ~ = [exp x~-, exp x~+] for any ~ • (0, 1]. (2.6)
(3) exp ~ e FR(Q) for every ~ ~ FR([2).

Proof. (1) Since )?(w) ~ o~o(R) for any w ~ O, and exp x (x e R) is a monotone continuous function, it follows
by Lemma 2.3 that
(exp ~)(w) = exp(£(w)) e ~-o(R) for each w • f2.
314 Qiao Zhong et al. / Fuzzy Sets and Systems 99 (1998) 311-324

(2) Observe the monotone continuity of exp x; by Lemma 2.3(1), we immediately obtain
(exp 2)~ = exp 2~ = [exp x~-, exp x ; ] for any a • (0, 1].
(3) For every 2 • FR(g2), using Theorem 1.2, we know that x~- and x ; are two random variables in
(O, d , P) for any a • (0, 1]. Since exp x (x • R) is continuous, then exp x~- and exp x ; are also random
variables, and it follows by Theorem 1.2 and (2) of this theorem that exp 2 • FR(f2). Hence proof of the
theorem is complete. []

Theorem 2.6. Let 2, 37e FR(f2). Then

exp(2 + 37) = (exp 2)(exp 37). (2.7)

Proof. Since for any 0~• (0, 1],

(2 + 37), = 2~ + 37, = [x~- + y~-, x ; + y ; ] ,
hence

(2 + 37)2 = x ; + y~-, (2 + 37); = x ; + y ; .

Thus, by Theorem 2.5 (2) we have
(exp(2 + y))~ = exp(2 + y)~

= [exp(2 + 37)2, exp(2 + 37);]

= [exp(x2 + Y2), exp(x; + y ; ) ]
= [(exp x~-) (exp y~-), (exp x ; ) (exp y ; ) ]
= [exp x~-, exp x~+] [exp y~-, exp y ; ]

= (exp 2)~ (exp)7)~

= (exp 2 exp 37)~ for any • • (0, 1-].

Consequently, exp(2 + 37) = (exp 2)(exp 37). []

Theorem 2.7. Let 2, 37• FR(f2);/f 2 ~< )7, then 0 <~ exp 2 ~< exp 37.

Proof. Let 2 ~< )7; then for any 0c• (0, 1], x - ~< y - , x + ~< y+.
Hence
0 ~< (exp 2)~ = [exp x~-, exp x;-]
~< [exp y~-, exp y;"] = (exp y)~.
This shows 0 ~< exp 2 ~ exp )7. []

2.2. Fuzzy random variable-valued exponential function a ~

Now we study the general fuzzy random variable-valued exponential function a ~, where a • R, a > 0 and
a~l.
 Qiao Zhong et al. / Fuzzy Sets and Systems 99 (1998) 311-324 315

Definition 2.8. Let a x (where a e R, a > 0 and a ¢ 1) be the o r d i n a r y exponential function to the base a. F o r
any ~ e FR(O), we define

a~ ~ ot{a~lxe;=}. (2.8)
a~(O, 1]

a ~ is called the fuzzy r a n d o m variable-valued exponential function to the base a.

T h e o r e m 2.9. a ~ has the following properties:

(1) (a~)(w) e ~ o ( R ) f o r every w e f2.

(2) I f a > 1, then
(a~)~ = [a ~;, a ~:] for any a e (0, 1]. (2.9)

(3) I f a < 1, then

(a~)~ = [a x;, a ~;] f o r any ct e (0, 1]. (2.10)

(4) a ~ e FR(t2) for every ~ e FR(I2).

Proof. (1) Observe that f(x) = a ~ (x e R) is continuous, and, by L e m m a 2.3, it follows that

(a~)(w) = a ~(w) e 3~:o(R) for every w e O.

(2) W h e n a > 1,f(x) = a * is a continuous m o n o t o n e increasing function, using L e m m a 2.3, it is easy to see
that
(a~)~ = { a ~ l x e ~ } = [a ~-, a x:] for any ct e (0, 1].

(3) W h e n a < 1,f(x) = a * is a continuous m o n o t o n e decreasing function; hence by using L e m m a 2.3, we

have

(a~)~ = {aXlx e ~ } = [a ~:, a ~-] for any ~ e (0, 1].

(4) F r o m (2) and (3) of this theorem, we m a y see that for any ct e (0, 1], (a~)~ is a closed r a n d o m interval
number. By T h e o r e m 1.2, this implies a ~ e FR(Y2) for each £ e FR(I2). Hence the p r o o f of the t h e o r e m is
complete. []

Theorem 2.10. L e t ~, )7 e FR(O). T h e n

(1) a ~/> 0.
(2) W h e n a > 1 , / f ~ ~< )7, then a ~ <<.a ~.
(3) W h e n a < 1, if ~ <~ )7, then a ~ >>.a;.

Proof. (1) F r o m T h e o r e m 2.9 we can see easily that for any ct e (0, 1], i f a > 1, then (a~)~ = [a ~:, a ~:] >t O, and
if a < 1, then (a~)~ = [a ~:, a ~;] >~ 0. Hence a ~ ~> 0.
(2) Consider the case a > 1. Let ~ ~< )7; then

x~-~<y~- and x~+~<y~+ for any ~ e (0,1].

It follows from T h e o r e m 2.9 that

(a~)~ = [a ~;, a ~:] <~ [a", a':] = (aY)~ for any ~ e (0, 1],
316 Qiao Zhong et al. / Fuzzy Sets and Systems 99 (1998) 311-324

so t h a t a ~ ,N<a y.
(3) W h e n a < 1, if 2 ~< 17, t h e n

x~-~<y~- and x~+ ~<y~ for a n y ~ s ( 0 , 1 ] .

By u s i n g T h e o r e m 2.9, we h a v e

(aY), = [a y:, a y;] <~ [ax:, a x:] = (a~), for a n y ~ e (0, 1].

C o n s e q u e n t l y , a y ~< a ~. []

Theorem 2.11. Let 2, 17 E FR(O). Then

(1) a ~+~ = a~.a y. (2.11)

(2) a ~ - -~ = a~/a ~. (2.12)

Proof. (1) Let a > 1, b y u s i n g T h e o r e m 2.9, for a n y ct e (0, 1]; we h a v e

(a ~ + Y,), = [a~ + m , a~ + ~):] = laX: +r,, aX,++y:] = laX;, a~,:] laY,, ay:]
= (a~), (aY), = (a ~. aY),;

hence, a ~ + ~ = a ~" a ~.
Let us c o n s i d e r a < 1. It follows f r o m T h e o r e m 2.9 that, for a n y ~ ~ (0, 1],

(a ~ + Y,)~ = [at~ + ~);, at~ + Y~);] = [aX: +r:, aX; +y;] = [aX;, aX;] [ay;, at,]

= (a~), (a~'), = ( d . aY),.

C o n s e q u e n t l y , a ~ + ~ = a ~- a ~.
(2) O b s e r v e first that, for a n y ~ e (0, 1],

(2 - y')~ = 2 , -- 37, = [x~- - Y2, x~+ - Y~-];

hence, (Y - y")~- = x~- - y~+ a n d (Y - y')2 = x~+ - y~-.

Let a > 1, f r o m T h e o r e m 2.9; we m a y see t h a t for a n y ~ s (0, 1],

(a ~-~), = [a~-;);, a~-~);] = [a x. -r:, aX:-r:] = [aX, / ar:, aX: /ar;]

= laX:, aX;]/[a y;, a r:] = (a~),/(a~), = (a~/a~),.

H e n c e we o b t a i n a ~ - ; = a~/a ~.
N e x t let us c o n s i d e r the case a < 1. F o r a n y ~ ~ (0, 1] it follows, b y T h e o r e m 2.9, t h a t

(a ~ - ~), = [a (~ - ~):, a ~ - ~),-] = [aX; -~;,a x. -~:] = [aX;/a ~-, a~;/a r:]

= [a~':, aX.]/[a ~:, a ~, ] = (a~),/(ag), = (ae/a~),,

so t h a t we also h a v e a ~ ; = a~/a ~. H e n c e , the p r o o f of the t h e o r e m is c o m p l e t e . []

R e m a r k . W h e n we c o n f i n e 2 ~ R, D e f i n i t i o n 2.8 is i d e n t i c a l w i t h the n o t i o n of the o r d i n a r y r e a l - v a l u e d

power function.
 Qiao Zhong et al. / Fuzzy Sets and Systems 99 (1998) 311-324 317

3. Fuzzy random variable-valued logarithmic function

In this section, we shall introduce the fuzzy random variable-valued logarithmic function using the
extension principle, and discuss some elementary properties for this kind of function.

Definition 3.1. Suppose that log` x (x e R) is the ordinary logarithmic function, where a ~ R, a > 0 and a ~ 1.
For any 9~e FR(f2) such that ~ > 0, we define

l°g. ~-zx U ~'{log`xlx~}. (3.1)

ae(O, 1]

log. ~ is called the fuzzy random variable-valued logarithmic function to the base a. If a = e, denote logo ~ by
In ~.

Theorem 3.2. log. ~ has the followin# properties:

(1) (log. x-)(w) ~ ~o(R) for every w ~ f2.
(2) I f a > 1, then

(log. ~)~ = [log` x~-, log. x ; ] for any ~ ~ (0, 1]. (3.2)
(3) I f a < 1, then

(log. ~)~ = [log` x~+, log. x~-] for any ~ ~ (0, 1]. (3.3)
(4) For every ~ ~ FR(f2) such that ~ > 0, log. ~ s FR(f2).

Proof. (1) Sincef(x) = log` x (x ~ R) is continuous, using Lemma 2.3 we immediately see that
(log` 2)(w) = log` ~(w) E ~o(R) for every w ~ f2.

(2) When a > 1, log` x is a continuous monotone increasing function. It follows by Lemma 2.3 that

(log. ~)= = {log. x lx ~ ~=} = [log` x~-, log` x~] for any • ~ (0, 1].

(3) When a < 1, log` x is a continuous monotone decreasing function. From Lemma 2.3, we obtain
(log. ~), = {log. x lx ~ ~,} = [log` x~+, log` x~-] for any ~ e (0, 1].

(4) From (2) and (3) of this theorem we may see that (log` ~), is a closed random interval number for any
e (0, 1]. By Theorem 1.2, log` ~ e FR(f2) for every ~ e FR(f2) such that Y > 0. Hence, we complete the proof
of the theorem. []

Theorem 3.3. Let 2, 37~ FR(I2) such that ,2 > 0 and 37 > O.
(1) When a > 1, ifY <~ 37, then log` Y ~< IOga 37.
(2) When a < 1, if ~ <<.37, then log` Y/> log~ 37.

Proof. (1) Consider the case a > 1. Suppose ~ ~< 37, then for any ~ e (0, 1] we have x~- ~ y~- and x~+ ~< y~+.
From Theorem 3.2, it follows that

(log` ~)~ = flog` x~, log` x + ]

~< [logo y~', log` y~] = (loga 37)~ for any ~ e (0, 1].
318 Qiao Zhong et al. / Fuzzy Sets and Systems 99 (1998) 311-324

This implies log. :~ ~< log. 37.

(2) Let a < 1. If £ ~< )7, then x~- ~< y~- and x~+ ~< y~+ for any ~ s (0, 1].
By T h e o r e m 3.2, we immediately obtain

(log. ~)~ = [log. x +, log, x~-]

i> [log. y+, log* y~-] = (log. 37)~ for any ~ ~ (0, 1].

T h a t is to say, log* ~ ~> log. 37. []

Theorem 3.4. log. ~ has the following properties:

(1) Let a > 1; when 0 < ~ < 1, then log. ~ < 0, and when 1 < ~, then log. ; > 0.
(2) Let a < 1; when 0 < ~ < 1, then log~ ~ > 0, and when 1 < ~, then log. £ < 0.

Proof. (1) Observe that when a > 1,f(x) = log* x (x e R) is a m o n o t o n e increasing function. Let 0 < £ < 1;
then for any a e (0, 1], we have 0 < [x~-, x~+] < 1, namely, 0 < x~- ~ x~+ < 1, hence

(log. ~)~ = [log, xd-, log. x ; ] < 0.

Thus, we obtain log* ~ < 0.

N o w consider the case 1 < ~. Since for any • E (0, 1], 1 < [x~-, x+], that is, 1 < x~- ~< x~+, it follows that

(log. ~)~ = [log. x~-, log. x~+] > 0.

This implies loga ~ > 0.

(2) Let a < 1. Under this c a s e f ( x ) = log* x (x ~ R) is a m o n o t o n e decreasing function. If 0 < ~ < 1, then
for any a e (0, 1], 0 < x~- ~< x~+ < 1; hence (log. ~)~ = [log* x +, log* x~-] > 0, and therefore we obtain
log. ~ > 0.
W h e n 1 < ~, for any ~ ~ (0, 1], we have 1 < x~- ~< x ; . It follows that

(log. ~)~ = [log, x ; , log. x~-] < 0 for any ~ ~ (0, 1].

Consequently, log* ~ < 0. Hence, the p r o o f of the theorem is complete. []

Theorem 3.5. Let 5c, 37 ~ FR(f2) such that ~ > 0 and 37 > O. Then

(1) log.(Y-37) = log* 2 + log* 37. (3.4)

(2) l o g , ( ; / ; ) = log. ~ - log. 37. (3.5)

Proof. (1) Consider the case a > 1. F o r any a ~ (0, 11, by using T h e o r e m 3.2, we have

(log.(~ • 37))~ = [log*(~ • 37)~-, 1Oga(-~"37)+]

[log*(xd- y~-), log*(x + y~+)]

[log~ x~- + log* y ; , log. x + + log. y+]

[IOga X~-, log* X+ ] + [log* y~-, log* y~+]

(log. ~)~ + (log* 37)~

(log. ~ + log* 37)~;

 Qiao Zhong et al. / Fuzzy Sets and Systems 99 (1998) 311-324 319

therefore, log`(2.37) = log` 2 + log. 37.

Furthermore, let a < 1; it follows by T h e o r e m 3.2 that

(log.(2"37)), = [1og.(2.)7)+, log`(2.37)~-]

= [log`(x + y~+), log.(x~- y~-)]

= [log. x~+ + log` y~+, log` x~- + log. y~-]

= [log` x~*, log` x~-] + [log` y~, log. y~-]

-- (log. 2), + (log. 37),

= (log. ; + log` 37), for any ~ e (0, 1].

Thus, we also obtain log.(2.37) = log. 2 + log. 37.

(2) If a > 1, applying T h e o r e m 3.2, then for any 0t e (0, 1],

(log.(2/37)), = [log`(2/37)~-, 1og.(2/37) +]

= [log`(x~-/y+~), log`(x~ +/yff)]

= [log. x~- -- log. y+, log. x + - log` y~-]

= [log` x~-, log. x + ] -- [log. y~-, log. y~+]

= (log. ~), - (log,, 37),

= (log` ~ - log` 37),,

which means that log`(2/37) = log` 2 - log` 37.

Moreover, let us consider the case a < 1. Using Theorem 3.2 we have

(log`(2/37)), = [log`(2/37) +, 1Oga(2/37)~-]

= [log`(x +/y~), log`(x~-/y])]

= [log` x g - log` y~-, log` x~- - log` y+]

= [log` x +, log. x~-] - [log. y+, log` y~-]

-- (log. ; ) , -- (log. 37),

-- (loga 9~ -- log` 37), for any • ~ (0, 1].

Thus, we also show that log`(2/37) = log` 2 - log. 37. Hence, the proof of the theorem is complete. []

The following theorem gives the relation between the fuzzy r a n d o m variable-valued exponential function
and logarithmic function.

Theorem 3.6. Suppose 2, 37 ~ FR(f2) such that 37 > O. T h e n a ~ = 37 if and o n l y / f 2 = log. 37, where a e R, a > 0
and a ~ 1.
320 Qiao Zhong et al. / Fuzzy Sets and Systems 99 (1998) 311-324

Pnmf. Necessity: Let a ~ = 37. If a > 1, then for any ~ e (0, 1], we know by T h e o r e m 2.9 that

Eye, y+] = 37, = (a~), = Ea x" , aX:],

and hence y~ = aX;, y+ = a x+", that is, x~- = log` y , , x + = log` y+. Thus, we have

2, = [x~-, x +] = [log, y~-, log` y+] = (log` 37), for any ~ e (0, 1],

which means that 2 = log, 37.

N o w let us consider the case a < 1. F r o m Theorem 2.9, it follows that

[y~, y+] = 37~ = (a~), = [a x:, a ~:] for any • s (0, 1].

Hence x~- = log,y +, x + = log, y~-, thus

2, = [x~-, x ~ ] = [log, y~+, log, y~-] = (log,)7), for any • ~ (0, 1].

We also obtain 2 = log, ~.

Observe that the above course may be inverted, hence the sufficiency of this theorem is also true. Hence we
complete the p r o o f of the theorem. []

Remark. If we confine 2 e R, then log, 2 degenerates into the ordinary logarithmic function.

Corollary 3.7. L e t n be any positive integer; then

log, 2" = n log, 2, (3.6)

where 2" = 2 ... £ (i.e. the product o f n 2's).

4. Fuzzy random variable-valued power function

In this section we shall introduce the concept of the fuzzy r a n d o m variable-valued power function, and
study its fundamental properties.

Definition 4.1. Let 2 e FR(f2) such that 2 ~> 0 or 2 < 0, and let r e R. We give the following definitions:
(1)
2°---1 i f 2 ~ 0. (4.1)

(2) If r is any positive integer, then

2'-2-.. 2 (the product of r £'s). (4.2)

(3) If r is any negative integer, then

2 ' - 1/2-'. (4.3)

(4) If r is any real number, then

2"zx U ~{x'lx~2~}, (4.4)

~(0,1]
 Qiao Zhong et al. / Fuzzy Sets and Systems 99 (1998) 311-324 321

w h e r e f ( x ) = x' (x • R) is the ordinary power function, and we require that ( - 1 ) ' - 1 is a real number when
~<0.

Theorem 4.2. £' has the following properties:

(1) (£')~ = { x ' l x • £ ~ } , f o r any ~ • ( 0 , 1]. (4.5)

(2) l f 2 > O, then for any e • (0, 1],

J'[(x2)', (x~+)r] /f r > 0, (4.6)

(~r)~ = [[(x2)', (x~-)'] if r < O.

(3) l f ~ < 0 (in this case, we require that ( - 1)'- 1 is a real number), then for any e • (0, 1],

)'[(x2)', (x~+)"] /f ( - 1 ) ' - 1 r > 0 (4.7)

(~')~ = ([(x:)', (x:)'] if ( - 1)'- 1 r < 0.

(4))?" • FR(f2).

Proof. (1) By L e m m a 2.3 we can immediately see that (1) of this theorem is true.
(2) Let £ > 0. We know that the ordinary power function f(x) = x r (where x • R, x > 0) is differentiable,
and

df(x) = r- x ' - 1;
dx

hence when r > 0, df(x)/dx > 0. It follows thatf(x) is a continuous monotone increasing function. Using (1)
of this theorem, we obtain

(~r)~ = [(x2)', (x2)'] for any ~ • (0, 1].

When r < 0, df(x)/dx < 0, hencef(x) is a continuous monotone decreasing function, by (1) of this theorem,
we also have

(~')~ = [(x2)', (x~-)'] for any ct • (0, 1].

(3) Let us consider the case ~ < 0. For the ordinary power functionf(x) = x r (where x • R, x < 0), since

df(x) _ r . x , _ 1 = ( _ 1 ) , _ l r ( _ x ) , _ l ,
dx

then when ( - 1)'- lr > 0, df(x)/dx > 0, hencef(x) is a continuous monotone increasing function. It follows by
(1) of this theorem that

(~r)~ = [(x~-y, (x~+)"] for any ~ • (0, 1].

When ( - 1)'- lr < 0, df(x)/dx < 0; thereforef(x) is a continuous monotone decreasing function, by (1) of
this theorem, we also obtain
(£')~ = [(x2) r, (x~-)'] for any • • (0, 1].

(4) (2) and (3) of this theorem tell us that (~')~ is a closed random interval number for any e • (0, 1]. It
follows by Theorem 1.2 that ~ • FR(O). Hence the proof of the theorem is complete. []
322 Qiao Zhong et al. / Fuzzy Sets and Systems 99 (1998) 311-324

Remark. In Definition 4.1, (2) and (3) coincide with (4). In fact, we have the following proof.
(1) Let r be any positive integer. Consider first 2 > 0.
On the one hand, by (2) of Definition 4.1, 2" denotes the product of r 2's; hence for any ~ • (0, 1],

(2'), = (2,)' = (x,+)q.

On the other hand, by (4) of Definition 4.1,

2"= U
• e(0, 13

Hence from Theorem 4.2 we also obtain

(•'), = { x ' t x • 2,} = [(x~-)', (x+) '] for any a • (0, 1],

which means that when 2 > 0, (2) is identical to (4) in Definition 4.1.
Next let us consider the case 2 < 0. When 2" is defined by (2) of Definition 4.1, for any a • (0, 1], we have

~'[(x+) ", (x~-)'] if r is even

(2'), = (2,)' = ~[(x~-)', (x~+)"] if r is odd.

When 2' is defined by (4) of Definition 4.1, by Theorem 4.2, we also obtain the same (2"),. Consequently, when
2 < 0, (2) is also identical to (4) in Definition 4.1.
(2) Let r be any negative integer. First let us consider the case 2 > 0. When 2' is defined by (3) of Definition
4.1, for any • • (0, 1], we have

(2"), = (1/2-'), = 1/[(x~-)-', (x+) - ' ] = [(x+) ', (x~-)'].

When 2' is defined by (4) of Definition 4.1, using Theorem 4.2 we also have

(2'), = [(x+) ", (x~-)'] for any • • (0, 1].

Hence, when 2 > 0, (3) is identical to (4) in Definition 4.1.

Furthermore, let us consider the case 2 < 0. When 2' is defined by (3) of Definition 4.1, for any ~ • (0, 1], it
follows that

(2'), = [[(x~-)', (x+) "] if r is even

)[(x+) ', (x~-)'] if r is odd.

When 2' is defined by (4) of Definition 4.1 we also obtain same (2'), for any ~ • (0, 1] by Theorem 4.2. That is
to say, (3) is identical to (4) in Definition 4.1 when 2 < 0.

Theorem 4.3. L e t 2 • FR(f2) such that 2 > O, r • R. T h e n

log~ 2" = r log~ 2, (4.8)

where a • R , a > 0 and a v~ 1.

Proof. It is clear that the equality (4.8) is true when r = 0.

(1) First consider the case r > 0. Under this case, by Theorem 4.2, for any • • (0, 1], we have

(~'), = [(x~-)', (x+)']. (4.9)

 Qiao Zhong et al. / Fuzzy Sets and Systems 99 (1998) 311-324 323

Let a > 1; it follows from T h e o r e m 3.2 and the equality (4.9) that

(log~ ~')~ = Elog~(~¢)~-, log~(~') +] = [log~(x~-)', log~(x~+)q -- [r log~ x~-, r log~ x + ]

= r(log~ ~)~ -- (r log~ ~)~ for any ct ~ (0, 1].

Thus, we obtain log~ ~r = r log~ ~.

If a < 1, by T h e o r e m 3.2 and equality (4.9), then for any a ~ (0, 1], we have

(log~ ~r)~ = [1oga(~r)+, Ioga(~r)~-] = [1og~(x+)r, Iog~(x;)r] = [r 1og~ x~+, r log~ x~-]

= r(log~ ~)~ = (r log~ ~)~ for any :t e (0, 1],

which implies log~ £r = r log~ ~.

(2) Let r < 0. By T h e o r e m 4.2, for any :t e (0, 1], we have

(~r)~ = [(x~)r, (x;)r]. (4.10)

W h e n a > 1, by T h e o r e m 3.2 and equality (4.10), it follows that

(log~ ~r)~ = [log~(~r)~-, log~(~)+] = [log~(x+)r, log,(x~-)r] = [r log, x~+, r log~ x~-]

= r[log~ x~-, logo x~+] = (r log~ ~)~ for any a e (0, 1].

Hence, logo ~r = r logo ~.

If a < 1, using T h e o r e m 3.2 and equality (4.10), we have

(log~ ~)~ = [log~(~r) f , log~(~r)~-] = [log~(x~-) r, loga(x~+) r] = [r log~ x~-, r log~ x~+ ]

= r[log~ x~+, log~ x~-] = (r log~ ~)~ for any a e (0, 1].

Therefore, we obtain log~ ~r = r log~ ~. Hence, p r o o f of the theorem is complete. []

Theorem 4.4. Let £¢ ~ FR(t~) such that ~ > O. Then

~r = exp(r In ~), (4.11)

where r ~ R.

Proof. By T h e o r e m 4.3 we know that

In ~r = r In ~.

It follows by T h e o r e m 3.6 that

~r = exp(r In ~). []

Theorem 4.5. Suppose ~, ~ ~ FR(12) such that ~ >1 0 and ~ >~ O, r ~ R.

(1) W h e n r > O, if ~ <~ ~, then ~r <~ ~r.
(2) W h e n r < O, if ~ <~ ~, then ~r >i fir.
324 Qiao Zhong et al. / Fuzzy Sets and Systems 99 (1998) 311-324

Proof. (1) Let 0 ~< £ ~< 37; then for any ~ e (0, 1], we have
0~<x~- ~<yd-, 0~<x~+ ~<y~+.
when r > 0, it follows by Theorem 4.2 that
(2')~ = [(x~-)', (x~+)'] ~< [( y~-)', (y~)'] = (37')~ for any c~~ (0, 1]; hence ~" ~< 37'.
(2) If r < 0, by Theorem 4.2, for any ct ~ (0, 1], we have
(~')~ = [(x,+) ", ( x ; ) q i> [( y+)', (y~-)'] = (7)~;
consequently, ~'/> 37r.
It is easy to see that £r is the ordinary power function when we confine ~ ~ R.
Applying fuzzy random variable-valued power functions, we may structure the following polynomial
function:
f(Y) = ~io + ~lY + a2 3~2 -F "'" "+" any n, (4.12)
where ~ FR(Q) (j = 1, 2, ... ,n) is a given fuzzy random variable, and Y ~ FR(f2) is an independent
variable.
Let

O(~) = ~;o + ~;~ + ~2~2 + ... + ~m~m (4.13)

be also a fuzzy random variable-valued polynomial function; we call

h(~) = f(~)/9(~) (4.14)

a fuzzy random variable-valued rational function.

References

['1] D. Dubois and H. Prade, Fuzzy Sets and Systems, Theory and Applications (Academic Press, New York, 1980).
[2] H. Kwakernaak, Fuzzy random variables, definitions and theorems, Inform. Sci. 15 (1978) 1-29.
[-3] R.E. Moore, Methods and Applications oflnterval Analysis (SIAM, Philadelphia, PA, 1979).
14] M.L. Purl and D.A. Ralescu, Fuzzy random variables, J. Math. Anal. Appl. 114 (1986) 409-422.
[5] Qiao Zhong, Theory of fuzzy random programming and fuzzy random optimum design of structures, Ph.D. Dissertation, Harbin
University of Architecture and Engineering (1994).
[6] Qiao Zhong, Zhang Yue and Wang Guangyuan, On fuzzy random linear programming, Fuzzy Sets and Systems 65 (1994) 31-49.
[-7] Qiao Zhong and Wang Guangyuan, On solutions and distribution problems of the linear programming with fuzzy random
variable coefficients, Fuzzy Sets and Systems 58 (1993) 155-170.
[-8] Qiao Zhong and Wang Guangyuan, Theory of Fuzzy Random Programming (Science Press, Beijing, 1996).
[9] Wang Guangyuan and Zhang Yue, The theory of fuzzy stochastic processes, Fuzzy Sets and Systems 51 (1992) 161-178.
[10] Wang Guangyuan and Qiao Zhong, Linear programming with fuzzy random variable coefficients, Fuzzy Sets and Systems 57
(1993) 295-311.
El 1] Wang Guangyuan and Qiao Zhong, Convergence of sequences of fuzzy random variables and its applications, Fuzzy Sets and
Systems 63 (1994) 187-199.
1-12] Zhang Yue and Wang Guangyuan, The Theory of Fuzzy Stochastic Dynamic Systems (Science Press, Beijing, 1993).