Journal of Applied Mathematics and Stochastic Analysis7, Number 2, 1994, 145-150
RANDOM APPROXIMATIONS ANDRANDOM FIXED POINT THEOREMS
ISMAT BEG1
Kuwait UniversityDepartment of Mathematics
P.O. Box 5969Safat 13060, KUWAIT
NASEER SHAHZADQuaid-i-Azarn University
Department o.f MathematicsIslamabad, PAKISTAN
(Received August, 1993; revised January, 1994)
ABSTRACT
Some results regarding random best approximation are proved. As ap-plications, several random fixed point theorems are obtained.
Key words: Random approximation, random fixed point, random operator,normed space.
AMS (MOS)subject classifications: 41A50, 47H40, 47H10, 60H25.
Random approximations and random fixed point theorems are stochastic generalizations of(classical) approximations and fixed point theorems. Random fixed point theory has receivedmuch attention for the last two decades, since the publication of the paper by Bharucha-Reid [3].On the other hand, random approximation has recently received further attention after the papersby Sehgal and Waters [11], Sehgal and Singh [10], Papageorgiou [9], Lin [8] and Beg and Shahzad[1]. The aim of this paper is to prove some theorems on random best approximations. Asapplications, some stochastic fixed point theorems are also derived.
Throughout this paper, (,A) denotes a measurable space and X be a Banach space. Let 2X
be the family of all subsets of X, CD(X) all nonempty closed subsets of X, CB(X) all nonemptybounded closed subsets of X and K(X) all nonempty compact subsets of X. A mappingF:f---,2X is called measurable if for any open subset C of X, F-I(C)-{wEf"F(w)NC : q}} E t. This type of measurability is usually called a weak measurability (cf.Himmelberg [6]). However, in this paper we always use this type of measurability, thus we omitthe term "weak" for simplicity. Let S be a nonempty subset of X, a map f: f S---,X is called a
random operator if for any x S, f(. ,x) is measurable. Similarly, a mapping F: S--CD(X)is a random multivalued operator if for every x S, F(.,x) is measurable. A measurablemapping :--S is called a measurable selector of the measurable mapping F:f--,CD(X) if
1On leave from the Department of Mathematics, Quaid-i-Azam University, Islamabad,Pakistan.Printed in the U.S.A. (C)1994 by North Atlantic Science Publishing Company 145
146 ISMAT BEG and NASEER SHAHZAD
((w) E F(w) for each w E f. A measurable mapping (: f2-*S is called a random fixed point of a
random operator f:fxS-*X (or F:2xS-*CD(X)) if for every w f, ((w)- f(w,((w)) (or((w) F(w,((w))). A map f’S-*X is called nonezpansive if [[ f(x)- f(y) I[ < [[ x- y II forx,y G S; f is a Banach operator of type k on S, if there exists a constant k, 0 < k < 1, such thatfor xS, [[f(x)-f2(x)[[ _<k [[x-f(x) l]; f is compact, if for any bounded subset S of X,f(S) (that is closure of f(S)) is compact. A random operator f:f2S-*X is continuous (non-expansive, Banach operator, compact etc.) if for each w E f2, f(w, is continuous (nonexpansive,Banach operator, compact etc.). For x X, define
and
yES
Ps(z) {y S: II - II I.
Here PS(x) denotes the set of all best approximations of y out of S and Ps(x), is called themetric projection on S. In the case when S is a Chebyshev set, Ps(z) is a point map X-,S, iscalled a proximity map. The set Ps(z) is closed and convex for any subspace (or convex subset)S of X. If a mapping f: X-*X leaves a subset M of X invariant, then a restriction of f to Mwill be denoted by the symbol f/M.
The following lemma is a consequence of [2, Theorem 2.1].Lemma 1" A continuous Banach random operator f:f2 x S-*S, where S is a separable closed
subset of a Banach space, has a random fixed point.
Remark 2: Lemma 1 remains true if S is a separable, closed subset of a normed space andf(o,S) is compact, for any w f.
The following is a random fixed point theorem for a nonexpansive random operator.corresponding fixed point theorem, we refer to Habinaik [5].
For a
Theorem 3: If f:f S-*S is nonexpansive random operator, where S is a separable closedand starshaped subset of a normed space E, and f(w,S) is compact, for any w [2, then f has arandom fixed point.
Proof: Let p be the starcenter of S and let {kn} be a sequence of positive numbers less than 1and converging to 1. For each n, define the random operator
fn(W, x) (1 kn)P + knf(W
For each n and w C fl, fn(w,’) maps S into itself because f(w,.):S-*S and S is starshaped.Moreover, each fn is a continuous Banach operator:
II (1-kn)P+knf(W,x)-(1-kn)P-knf(W,(1-kn)P+knf(W,x))[I
kn II f(w, z) f(co, (1 kn)p + knf(W x)) II_< k, II - fn(w, x)II, for each
Since f(w,S) compact, fn(w,S) is compact too. By Remark 2, fn has a random fixed point n"For each n, define Gn:f-*tt’(S as Gn(w {i(w)’i>n}. Define G:f2-*K(S) by G(w)
Random Approximations and Random Fixed Point Theorems 147
Gn(w). Then, G is measurable [5] and has a measurable selector (. Thus, ( is a desiredn----1
random fixed point of f. Fix any w E ft. Since f(w,S) is compact, {f(w,n(w))} has a
subsequence {f(w,n (w))} convergent to (w); since kn --1, the subsequencem m
nm(W)-(1-knm)P+knmf(W,nm(W)) converges to (w). By continuity of f, f(W,nm(W))converges to f(w, (w)). Thus f(, ()) ().
Theorem 4: Let E be a normed space. If f’x EE is a nonexpansive random operatorwith deterministic fixed point x, leaving a subspace M of E invariant, f(w,. )/M is compact andPM(X) is separable, then the point x has a best random approximation ’f--.M which is also arandom fixed point of f.
Proof: The set PM(X) is invariant with respect to the operator f(w,.) for eachIndeed, from [12] it follows that there exists b PM(X), that is, d(x,M)- II - b II. Since f isnonexpansive and x- f(w,x), we have
d(x, M) II - b II >_ II - f(w, b)So, II - f(,b)II < d(,M) for each b PM(X). Therefore, PM(X) is f(w,. )-invariant,
that is, f(W, PM(X)) C PM(X) for each w f. The set PM(Z) is closed and starshaped since it is
convex; since PM(Z) is a bounded subset of M and f(w,. )/M is compact, the set f(W, PM(Z)) iscompact, for any w f.
Theorem 3 further implies that f has a random fixed point in subset PM(X). Hence the ran-dom operator f has a random fixed point in the set of best approximation PM(X).
Theorem 5: Let f:fxXX be a nonezpansive random operator with deterministic fixedpoint x, leaving convex subset S of X invariant and f(w,.)/S compact. If Ps(x) is nonemptyand separable, then point x has a best random approximation :f--S which is also a randomfixed point of f.
Proof: Proof is exactly same as that of Theorem 4; instead of Theorem 3, we use Itoh [7,Corollary 2.2] to show that f has a random fixed point.
Theorem 6: Let S be a nonempty separable closed convex subset of a Hilbert space H andf:fS---,H a nonexpansive random operator with f(w,S) compact for any w f. Then thereexists a measurable map :---,S such that
II e()- f(w,,(w))II d(f(w,,(w)), S)
for each w f.
Proof: Let P: H---S be a proximity map. From [3] we have that, P is nonexpansive in the
Hilbert space. Then P o f: f S---S is also a nonexpansive random operator. Now P o f(w,S) is
compact since P o f(w, S) C P o f(w, S) and P o f(w, S) is compact, for any w Ft. By Theorem3, there exists a measurable map :-S such that P o f(w, (w)) = (w) for each w 6 f. Hence,
II 5()- f(w, ,())II II P o f(, ,())- f(, ())II
=d(f(w,(w)),S),for each
Theorem 7: Let S be a nonempty separable closed convex subset of a Hilbert space H, and
f:f2xSH be a nonexpansive random operator with f(w,S) compact for any w f2, such that
148 ISMAT BEG and NASEER SHAHZAD
for each x on the boundary of , II f(w,x)- l(W)II _< II -q()II fo om measurable map
l.f2S. Then f has a random .fixed point.
Proof: Theorem 6 implies that there exists a measurable map : flS such that
If for any w ( f2, f(w,(w))(5 S then ((w)- f(w, ff(w)). Suppose f(w,(w)) S. Let P be theproximity map on S, then P is nonexpansive. Therefore P o f: f x SS is nonexpansive randomoperator. By definition
l[ f(w, (()) P o f(, ((w)) II d(f(w, (()), S).
It implies that P o f(w,(w))- ((w) for each w ( n. By hypothesis, there is a measurable map
(l:flS such that II f(,())-C()II < II ()-x() II, Now it suffices to show that(w) l(w) for each w ft. Otherwise, we have
It follows that there is a measurable map 7: flS defined by
7(w) a(w) + (1 o)1(o),
with 0 < a[ < 1 for which
II 7(w) f(w, ((w))II < II ()- f(w, (w))II.
This is a contradiction. Hence (w) l(W) for each w ( f.
Theorem 8: Let S be a nonempty separable closed convex subset of a Hilbert space H. Letf:f x S---H be nonexpansive random operator with f(w,S) compact for any w E f, and assumethat for any measurable map l’-OS (where OS denotes the boundary of S), with(l(w)- P o f(w,(l(W)) that is a random fixed point of Po f, then f has a random fixed point.
Proof: From Theorem 6, there exists measurable map (" f--.S such that
II ()- f(, ())II d(f(w,(w)),S).
If for any w Efl, f(w, ((w)) e S then f(w, (()) ((w). Suppose f(w, ((w)) S. As in Theorem7 we have P o f(w,((w))- ((w) for each 0 ft. If (:flOS, then by the above assumption,f(w, (w)) ((w) for each w ft. Otherwise, f(w, ((w)) S. Thus for f, ((w) P o f(w, (w))
A random operator f’f2 x SH is said to satisfy property (I), if f(w,z)- t(w)z for somein cOS, then c(w) < 1 for each w 6 f. [nr : --.a i measurable map.]
Theorem 9: Let S be a separable closed ball centered at the origin and with radius r in aHilbert space H. Let f:f x SH be a nonexpansive random operator with compact f(w,S) forany w f and satifying property (I). Then, f has a random fixed point.
Proof: There exists a measurable map (: fl---,S such that
II ff(.) f(w, ((w)) II d(f(w, ()), S).
If for any w f, f(w, ((w)) S then f(w, ((w)) ((w). Suppose II f(, (())II > . By con-
Random Approximations and Random Fixed Point Theorems 149
vexity, ((w) is on the boundary of S and f(w,((w))= a(w)((co) for some measurable mapa: f-(O, c). It gives for each
.() II-()() II II f(, ()) II r> -1.
Hence, this is a contradiction to property (I).Theorem 10: Let S be a nonempty separable closed convex subset of a Hilbert space H and
f:f x S---+H be a nonexpansive random operator with compact f(co, S), for any co f. Moreover,f satisfies one of the following conditions:
(i)
(ii)
For each co f, each z S with :c 5 f(w,x), there exists V, dependenton w and z, in Is(z) = {z + c(z- x): some zeS, c>O} such that
II y- f(,x)II < II - f(w,x)II.f is weakly inward (that is, for each w , f(w,x) 6 Is(z for any x S).
Then, f has a random fixed point.
For the proof, see Theorem 6 and Lin [8, Proof of Theorem 4].
Pferences
[1] Beg, I. and Shahzad, N., Random fixed points of random multivalued operators onPolish space, Nonlinear Anal. 21} (1993), 835-847.
[21 Beg, I. and Shahzad, N., Random fixed points and approximations in random convexmetric spaces, J. Appl. Math. Stoch. Anal. 6, No. 3 (1993), 237-246.
[3] Bharucha-Reid, A.T., Fixed point theorems in probabilistic analysis, Bull. Amer. Math.Soc. 82 (1976), 641-657.
[4] Cheney, W. and Goldstein, A.A., Proximity maps for convex sets, Proc. Amer. Math.Soc. 10 (1959), 448-550.
[5] Habiniak, L, Fixed point theorems and invariant approximations, J. Approx. Theory 56(1989), 241-244.
[6] Himmelberg, C.J., Measurable relations, Fund. Math. 87 (1975), 53-72.
[7] Itoh, S., Random fixed point theorems with an application to random differentialequations in Banach spaces, J. Math. Anal. Appl. 67 (1979), 261-273.
[8] Lin, T.C., Random approximations and random fixed point theorems for non-self-maps,Proc. Amer. Math. Soc. 103 (1988), 1129-1135.
[9] Papageorgiou, N.S., Random fixed point theorems for measurable multifunctions inBanach spaces, Proc. Amer. Math. Soc. 97 (1986), 507-514.
[10] Sehgal, V.M. and Singh, S.P., On random approximations and a random fixed pointtheorem for set valued mappings, Proc. Amer. Math. Soc. 95 (1985), 91-94.
150 ISMAT BEG and NASEER SHAHZAD
[11] Sehgal, V.M. and Waters, C., Some random fixed point theorems for condensingoperators, Proc. Amer. Math. Soc. 90 (1984), 425-429.
[12] Smoluk, A., Invariant approximations, Mathematyka Stosowana 17 (1981), 17-22[Polish].
Submit your manuscripts athttp://www.hindawi.com
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttp://www.hindawi.com
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Journal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
CombinatoricsHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
International Journal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
The Scientific World JournalHindawi Publishing Corporation http://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com
Volume 2014 Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Stochastic AnalysisInternational Journal of