We present some results concerning the extension of a semi-Hölder real-valued function defined on a subset of a quasi-metric space, preserving some shape properties: the smallest semi-Hölder constant, the radiantness and the global minimum (maximum) of the extended function.
Authors
Costică Mustăța ”Tiberiu Popoviciu” Institute of Numerical Analysis Cluj-Napoca, Romanian Academy
Keywords
Quasi-metric space; semi-Holder function; radiant function
Paper coordinates
Mustăţa, C., On the extensions preserving the shape of a semi-Hölder function, Results. Math. 63 (2013), 425–433.
doi: 10.1007/s00025-011-0206-x
[1] Breckner, W.W.: Holder continuity of certain generalized convex functions. Optimization 28, 201–209 (1994)
[2] Cobzas, S.: Functional analysis in asymmetric normed spaces, arXiv:1006. 1175v1.
[3] Collins, J., Zimmer, J.: An asymmetric Arzel`a-Ascoli theorem. Topol. Appl. 154, 2312–2322 (2007)
[4] Doagooei, A.R., Mohebi, H.: Optimization of the difference of ICR functions. Nonlinear Anal. Theor. Method Appl. 71, 4493–4499 (2009)
[5] Flecther, P., Lindgren, W.F.: Quasi-uniform spaces. Dekker, New York (1982)
[6] Krein, M.G., Nudel’man, A.A.: The Markov moment problem and extremum problems, Nauka, Moscow 1973 (in Russian), English translation . American Mathematical Society, Providence (1977)
[7] Kunzi, H.-P.A.: Nonsymmetric distances and their associated topologies: about the origin of basic ideas in the area of asymmetric topology. In: Aull, C.E., Lowen, R. Handbook of the history of general topology, pp. 853–968. Kluwer Academic Publishers, Dordrecht (2001)
[8] Martinez-Legaz, J.E., Rubinov, A.M., Schaible, S.: Increasing quasi-concave coradiant functions with applications in mathematical economics. Math. Oper. Res. 61, 261–280 (2005)
[9] Matouskova, E.: Extensions of continuous and Lipschitz functions. Canad. Math. Bull. 43(2), 208–217 (2000)
[10] McShane, E.T.: Extension of range of functions. Bull. Am. Math. Soc. 40, 837–842 (1934)
[11] Mennucci, A.: On asymmetric distances, Tech. Rep. Scuola Normale Superiore, Pisa (2004)
[12] Mustata, C.: Extension of semi-Lipschitz functions on quasi-metric spaces. Rev. Anal. Num´er. Th´eor. Approx. 30, 61–67 (2001)
[13] Mustata, C.: A Phelps type theorem for spaces with asymmetric norms. Bull. Stiint¸. Univ. Baia Mare, Ser. B. Matematic˘a-Informatic˘a 18, 275–280 (2002)
[14] Mustata, C.: On the approximation of the global extremum of a semi-Lipschitz function. Mediterr. J. Math. 6, 169–180 (2009)
[15] Mustata, C.: Extensions of semi-H¨older real valued functions on a quasi-metric space. Rev. Anal. Num´er. Th´eor. Approx. 38, 164–169 (2009)
[16] Phelps, R.R.: Uniqueness of Hahn-Banach extension and unique best approximation. Trans. Am. Math. Soc. 95, 238–255 (1960)
[17] Romaguera, S., Sanchis, M.: Semi-Lipschitz functions and best approximation in quasi-metric spaces. J. Approx. Theor. 103, 292–301 (2000)
[18] Romaguera, S., Sanchis, M.: Properties of the normed cone of semi-Lipschitz functions. Acta Math. Hungar 108, 55–70 (2005)
Paper (preprint) in HTML form
2013-Mustata-On the Extensions Preserving-Result Math
On the Extensions Preserving the Shape of a Semi-Hölder Function
Costică Mustăţa
Abstract
We present some results concerning the extension of a semiHölder real-valued function defined on a subset of a quasi-metric space, preserving some shape properties: the smallest semi-Hölder constant, the radiantness and the global minimum (maximum) of the extended function.
Let XX be a nonvoid set. A mapping d:X xx X rarr[0,oo)d: X \times X \rightarrow[0, \infty) satisfying the following conditions: (QM_(1))d(x,y)=d(y,x)=0quad\left(\mathrm{QM}_{1}\right) d(x, y)=d(y, x)=0 \quad iff x=yx=y, (QM_(2))d(x,y) <= d(x,z)+d(z,y)\left(\mathrm{QM}_{2}\right) d(x, y) \leq d(x, z)+d(z, y),
for all x,y,z in Xx, y, z \in X is called a quasi-metric (asymmetric metric) on XX, and the pair ( X,dX, d ) is called a quasi-metric space [ 17,18 ].
Because, in general, d(x,y)!=d(y,x),x,y in Xd(x, y) \neq d(y, x), x, y \in X, one defines the conjugate bar(d)\bar{d} of quasi-metric dd as the quasi-metric bar(d)(x,y)=d(y,x),x,y in X\bar{d}(x, y)=d(y, x), x, y \in X.
For example, an asymmetric norm ||∣\| \mid on a linear space XX (see [6], Ch. IX, § 5) or [2], where a functional analysis in asymmetric normed space is presented) defines a quasi-metric d_(||∣)d_{\| \mid}through the formula:
d_(||)|(x,y)=||y-x|,x,y in X.d_{\|}|(x, y)=\| y-x|, x, y \in X .
Let ( X,dX, d ) be a quasi-metric space. A sequence (x_(k))_(k >= 1)\left(x_{k}\right)_{k \geq 1} is dd-convergent to x_(0)in Xx_{0} \in X (or forward convergent to x_(0)in Xx_{0} \in X ) if
We say that the set Y sub XY \subset X is dd-closed ( bar(d)\bar{d}-closed) if every dd-convergent ( bar(d)\bar{d}-convergent) sequence (y_(n))_(n >= 1)sub Y\left(y_{n}\right)_{n \geq 1} \subset Y has limit in YY.
We say that a set Y sub XY \subset X is dd-sequentially compact (forward sequentially compact) if every sequence in YY has a dd-convergent (forward convergent) subsequence with limit in YY (Definition 4.1 in [3]). Finally, the set YY in (X,d)(X, d) is called ( d, bar(d)d, \bar{d} )-sequentially compact if every sequence (y_(n))_(n >= 1)\left(y_{n}\right)_{n \geq 1} in YY has a subsequence (y_(n_(k)))_(k >= 1)d\left(y_{n_{k}}\right)_{k \geq 1} d-convergent to u in Yu \in Y and bar(d)\bar{d}-convergent to v in Yv \in Y. For other properties and results in asymmetric metric spaces, see also [3,5,7,11-18].
2. Extension of Semi-Hölder Functions
Let ( X,dX, d ) be a quasi-metric space, Y sub XY \subset X be a nonvoid subset of XX and alpha in(0,1]\alpha \in(0,1] a given number.
Definition 1. A function f:Y rarrRf: Y \rightarrow \mathbb{R} is called dd-semi-Hölder (of exponent alpha\alpha ) if there exists a constant K_(Y)(f) >= 0K_{Y}(f) \geq 0 such that
for all x,y in Yx, y \in Y.
For ff a function dd-semi-Hölder on YY denote by
{:(2)||f|_(Y,d)^(alpha):=s u p{((f(x)-f(y))vv0)/(d^(alpha)(x,y)):d(x,y) > 0,x,y in Y}:}\begin{equation*}
\|\left. f\right|_{Y, d} ^{\alpha}:=\sup \left\{\frac{(f(x)-f(y)) \vee 0}{d^{\alpha}(x, y)}: d(x, y)>0, x, y \in Y\right\} \tag{2}
\end{equation*}
the smallest constant K_(Y)(f)K_{Y}(f), satisfying the inequality (1).
A function f:Y rarrRf: Y \rightarrow \mathbb{R} is called <= _(d)\leq_{d}-increasing if f(x) <= f(y)f(x) \leq f(y) whenever d(x,y)=0,x,y in Yd(x, y)=0, x, y \in Y.
The set R_( <= _(d))^(Y)\mathbb{R}_{\leq_{d}}^{Y} of all <= _(d)\leq_{d}-increasing functions on YY is a cone in the linear space R^(Y)\mathbb{R}^{Y} of all real-valued functions defined on YY.
One denotes by
{:(3)Lambda_(alpha)(Y","d):={f inR_( <= _(d))^(Y)∣f" is "d"-semi-Hölder "}:}\begin{equation*}
\Lambda_{\alpha}(Y, d):=\left\{f \in \mathbb{R}_{\leq_{d}}^{Y} \mid f \text { is } d \text {-semi-Hölder }\right\} \tag{3}
\end{equation*}ö
the set of all dd-semi-Hölder functions on YY. This set is a subcone of the cone R_( <= _(d))^(Y)\mathbb{R}_{\leq_{d}}^{Y}.
The functional |||_(Y,d)^(alpha):Lambda_(alpha,0)(Y,d)rarr[0,oo)\|\left.\right|_{Y, d} ^{\alpha}: \Lambda_{\alpha, 0}(Y, d) \rightarrow[0, \infty) defined by (2) satisfies the axioms of an asymmetric norm, and ( Lambda_(alpha,0)(Y,d),|||_(Y,d)^(alpha)\Lambda_{\alpha, 0}(Y, d), \|\left.\right|_{Y, d} ^{\alpha} ) is an asymmetric normed cone (compare with [18]).
Observe that f inLambda_(alpha)(Y,d)f \in \Lambda_{\alpha}(Y, d) if and only if -f inLambda_(alpha)(Y, bar(d))-f \in \Lambda_{\alpha}(Y, \bar{d}); moreover ||f|_(Y,d)^(alpha)=||-f|_(Y, bar(d))^(alpha)\left\|\left.f\right|_{Y, d} ^{\alpha}=\right\|-\left.f\right|_{Y, \bar{d}} ^{\alpha}.
Example 1. Let YY be a set in a quasi-metric space ( X,dX, d ) and let y_(0)in Yy_{0} \in Y be fixed. For a number alpha in(0,1]\alpha \in(0,1] one considers the function f:Y rarrR,f(y)=d^(alpha)(y,y_(0))f: Y \rightarrow \mathbb{R}, f(y)= d^{\alpha}\left(y, y_{0}\right). Then f inLambda_(alpha,0)(Y,d)f \in \Lambda_{\alpha, 0}(Y, d). Indeed, for all y_(1),y_(2)in Yy_{1}, y_{2} \in Y,
The last inequality follows by the following simple lemma:
Lemma 1. Let a,b,ca, b, c be real nonnegative numbers such that a <= b+ca \leq b+c. Then for alpha in(0,1]\alpha \in(0,1] it follows a^(alpha) <= b^(alpha)+c^(alpha)a^{\alpha} \leq b^{\alpha}+c^{\alpha}.
Example 2. Let ( X,||∣X, \| \mid ) be an asymmetric normed space. For a fixed y_(0)in Xy_{0} \in X and alpha in(0,1]\alpha \in(0,1] the function h(x)=||x-y_(0)|^(alpha)h(x)=\| x-\left.y_{0}\right|^{\alpha} is d_(||)d_{\|}-semi-Hölder, where {:d_(||)∣y_(0),x)=||x-y_(0)∣,x in X\left.d_{\|} \mid y_{0}, x\right)=\| x-y_{0} \mid, x \in X. Using Lemma 1 it follows h(x_(1))-h(x_(2)) <= ||x_(1)-x_(2)|^(alpha),x_(1),x_(2)in Xh\left(x_{1}\right)-h\left(x_{2}\right) \leq \| x_{1}-\left.x_{2}\right|^{\alpha}, x_{1}, x_{2} \in X, and h(y_(0))=0h\left(y_{0}\right)=0. This means that h inLambda_(alpha,0)(X,d_(||))h \in \Lambda_{\alpha, 0}\left(X, d_{\|}\right).
Remark 1. By Lemma 1 it follows that if dd is a quasi-metric on XX, then d^(alpha)(alpha in(0,1])d^{\alpha}(\alpha \in(0,1]) is also a quasi-metric on XX. In fact a dd-semi-Hölder function ff (of exponent alpha in(0,1])\alpha \in(0,1]) on YY is a d^(alpha)d^{\alpha}-semi-Lipschitz function on ( Y,d^(alpha)Y, d^{\alpha} ) (see [17], for the definition of semi-Lipschitz functions).
The following theorem holds.
Theorem 1. Let ( X,dX, d ) be a quasi-metric space, let YY be a nonvoid subset of XX, let alpha in(0,1]\alpha \in(0,1], and let f inLambda_(alpha)(Y,d)f \in \Lambda_{\alpha}(Y, d). Further, let E_(d)(f)\mathcal{E}_{d}(f) be defined by
whenever x in Xx \in X.
Proof. If f inLambda_(alpha)(Y,d)f \in \Lambda_{\alpha}(Y, d) then ff is d^(alpha)d^{\alpha}-semi-Lipschitz on (Y,d^(alpha))\left(Y, d^{\alpha}\right). By ([12], Theorem 2) it follows that the functions F_(d)(f)F_{d}(f) defined by (6), and G_(d)(f)G_{d}(f) defined by (7) satisfy
Consequently, the statements 1^(0)1^{0} and 2^(0)2^{0} are proved.
The inequalities (8) are proved in [12] for semi-Lipschitz functions (see also ([15], Remark 3), so that the statement 3^(0)3^{0} holds too.
The set E_(d)(f)\mathcal{E}_{d}(f) defined in (5) is called the set of extensions of f inLambda_(alpha)(Y,d)f \in \Lambda_{\alpha}(Y, d) (preserving the smallest constant ||f|_(Y,d)^(alpha)\|\left. f\right|_{Y, d} ^{\alpha} ). The functions F_(d)(f)F_{d}(f), respectively G_(d)(f)G_{d}(f) are called the maximal extension, respectively the minimal extension of ff [see (8)].
Remark 2. In [15] one gives a direct proof of Theorem 1, by considering the function G_(d)(f)G_{d}(f) defined by (7) and proving that G_(d)(f)G_{d}(f) is well defined, G_(d)(f)|_(Y)=f\left.G_{d}(f)\right|_{Y}=f and ||G_(d)(f)|_(X,d)^(alpha)=||f|_(Y,d)^(alpha)\left.\left\|\left.G_{d}(f)\right|_{X, d} ^{\alpha}=\right\| f\right|_{Y, d} ^{\alpha} (see also [10,12]).
A natural problem is the following: If f inLambda_(alpha)(Y,d)f \in \Lambda_{\alpha}(Y, d) has some supplementtary properties, does there exist H inE_(d)(f)H \in \mathcal{E}_{d}(f) preserving these properties? Such a problem is considered in [9] for Lipschitz functions.
We shall consider two problems of such kind.
For the first one, in the sequel ( X,dX, d ) is a quasi-metric linear space and Y sub XY \subset X is a subset of XX.
The set YY is said to be radiant if it has the following properties:
(i) YY is nonvoid;
(ii) lambda y in Y\lambda y \in Y for all y in Yy \in Y and all lambda in[0,1]\lambda \in[0,1].
Let YY be a radiant set in XX, and let f:Y rarrRf: Y \rightarrow \mathbb{R}, and let alpha in(0,1]\alpha \in(0,1]. The function ff is said to be alpha\alpha-radiant if
for all y in Yy \in Y and all lambda in(0,1]\lambda \in(0,1].
The 1 -radiant functions are called, simply, radiant.
Observe that all radiant sets in a linear space XX contain the null element theta\theta of XX, and every alpha\alpha-radiant function satisfies f(theta) <= 0f(\theta) \leq 0. We consider only functions satisfying f(theta)=0f(\theta)=0.
The function f:Y rarrRf: Y \rightarrow \mathbb{R} is said to be alpha\alpha-co-radiant ( alpha in(0,1]\alpha \in(0,1] ) if
for all x,y in Yx, y \in Y and lambda in[0,1]\lambda \in[0,1] (see [1]).
If theta in Y\theta \in Y and YY is convex, then every convex function ( alpha=1\alpha=1 ) on YY with f(theta)=0f(\theta)=0 is radiant, and every alpha\alpha-convex function on YY such that f(theta)=0f(\theta)=0 is alpha\alpha-radiant, because
for all x,y in Xx, y \in X and lambda >= 0\lambda \geq 0. Such a quasi-metric is for example d_(||)∣d_{\|} \mid, generated by an asymmetric norm ||∣\| \mid.
The following result holds.
Theorem 2. Let ( X,dX, d ) be a quasi-metric linear space with dd positively homogeneous, let YY be a radiant subset of XX, let alpha in(0,1]\alpha \in(0,1] and let f inLambda_(alpha)(Y,d)f \in \Lambda_{\alpha}(Y, d). Then the following statements hold: 1^(0)1^{0} If ff is alpha\alpha-radiant, then F_(d)(f)F_{d}(f) is alpha\alpha-radiant. 2^(0)2^{0} If ff is alpha\alpha-co-radiant, then G_(d)(f)G_{d}(f) is alpha\alpha-co-radiant. 3^(0)3^{0} If ff is inverse co-radiant, then F_(d)(f)F_{d}(f) is inverse co-radiant.
Proof. Let f:Y rarrRf: Y \rightarrow \mathbb{R} be radiant, and f inLambda_(alpha)(Y,d)f \in \Lambda_{\alpha}(Y, d). Let us consider the maximal extension F_(d)(f)F_{d}(f). Then for all lambda in[0,1]\lambda \in[0,1] and y in Yy \in Y,
for every x in Xx \in X, showing that F_(d)(f)F_{d}(f) is alpha\alpha-radiant.
Now, let x in X,lambda in[0,1]x \in X, \lambda \in[0,1], and f inLambda_(alpha)(Y,d)f \in \Lambda_{\alpha}(Y, d) be alpha\alpha-co-radiant. Then, for every y in Yy \in Y, by considering the minimal extension G_(d)(f)inE_(d)(f)G_{d}(f) \in \mathcal{E}_{d}(f) one gets:
Taking the supremum with respect to y in Yy \in Y, one obtains
G_(d)(f)(lambda x) >= lambda^(alpha)G_(d)(f)(x),x in XG_{d}(f)(\lambda x) \geq \lambda^{\alpha} G_{d}(f)(x), x \in X
and the statement 2^(0)2^{0} is proved.
Finally, if f inLambda_(alpha)(Y,d)f \in \Lambda_{\alpha}(Y, d) is inverse co-radiant on the radiant set YY, then for every y in Yy \in Y and lambda in(0,1]\lambda \in(0,1], and for the maximal extension F_(d)(f)F_{d}(f) one obtains:
Taking the infimum with respect to y in Yy \in Y it follows
F_(d)(f)(lambda x) <= (1)/(lambda)F_(d)(f)(x),x in X,F_{d}(f)(\lambda x) \leq \frac{1}{\lambda} F_{d}(f)(x), x \in X,
and the statement 3^(0)3^{0} holds.
Another property preserved by extensions is the global minimum (maximum) of a function f inLambda_(alpha)(Y,d)f \in \Lambda_{\alpha}(Y, d).
Let ( X,dX, d ) be a quasi-metric space, and let Y sub XY \subset X be a nonempty subset of XX. An element y_(0)in Yy_{0} \in Y is called a global minimum (maximum) point of f inLambda_(alpha)(Y,d)f \in \Lambda_{\alpha}(Y, d) if
for all y in Yy \in Y.
Theorem 3. Let ( X,dX, d ) be a quasi-metric space, let YY be a nonvoid subset of XX, let y_(0)in Yy_{0} \in Y, let alpha in(0,1]\alpha \in(0,1], and let f inLambda_(alpha)(Y,d)f \in \Lambda_{\alpha}(Y, d). Then the following statements hold: 1^(0)1^{0} If YY is d-closed, then y_(0)in Yy_{0} \in Y is a global minimum point for ff in YY if and only if y_(0)y_{0} is a global minimum point of F_(d)(f)F_{d}(f) in XX. 2^(0)2^{0} If YY is bar(d)\bar{d}-closed, then y_(0)in Yy_{0} \in Y is a global maximum point for ff in YY if and only if y_(0)y_{0} is a global maximum point of G_(d)(f)G_{d}(f) in XX.
Proof. quad1^(0)\quad 1^{0} Let y_(0)in Yy_{0} \in Y be a global minimum point of f inLambda_(alpha)(Y,d)f \in \Lambda_{\alpha}(Y, d). For every y in Yy \in Y we have
If x!in Y,Yx \notin Y, Y being dd-closed, there exists delta > 0\delta>0 such that d(x,y) > deltad(x, y)>\delta for all y in Yy \in Y. Consequently,
{:[F_(d)(f)(x)=i n f_(y in Y){f(y)+||f|_(Y,d)^(alpha)d^(alpha)(x,y)}],[ >= i n f_(y in Y){f(y)+||f|_(Y,d)^(alpha)delta^(alpha)}],[=f(y_(0))+||f|_(Y,d)^(alpha)delta^(alpha) >= f(y_(0))","]:}\begin{aligned}
F_{d}(f)(x) & =\inf _{y \in Y}\left\{f(y)+\|\left. f\right|_{Y, d} ^{\alpha} d^{\alpha}(x, y)\right\} \\
& \geq \inf _{y \in Y}\left\{f(y)+\|\left. f\right|_{Y, d} ^{\alpha} \delta^{\alpha}\right\} \\
& =f\left(y_{0}\right)+\|\left. f\right|_{Y, d} ^{\alpha} \delta^{\alpha} \geq f\left(y_{0}\right),
\end{aligned}
so that for every x in X,F_(d)(f)(x) >= f(y_(0))=F_(d)(f)(y_(0))x \in X, F_{d}(f)(x) \geq f\left(y_{0}\right)=F_{d}(f)\left(y_{0}\right).
Conversely, suppose that y_(0)in Xy_{0} \in X is a global minimum point for F_(d)(f)F_{d}(f) in XX. If we would show that y_(0)in Yy_{0} \in Y, then, as F_(d)(f)|_(Y)=f\left.F_{d}(f)\right|_{Y}=f, it would follow that y_(0)y_{0} is a global minimum point for ff in YY.
Case I: ||f|_(Y,d)^(alpha)=0\|\left. f\right|_{Y, d} ^{\alpha}=0.
In this case there exists c inRc \in \mathbb{R} such that f(y)=cf(y)=c for all y in Yy \in Y. It follows ||F_(d)(f)|_(X,d)^(alpha)=0\|\left. F_{d}(f)\right|_{X, d} ^{\alpha}=0, so that F_(d)(f)=F_{d}(f)= const on XX. Since F|_(Y)=f\left.F\right|_{Y}=f we must have F_(d)(f)(x)=F_{d}(f)(x)= const , for all x in Xx \in X.
Case II: ||f|_(Y,d)^(alpha) > 0\|\left. f\right|_{Y, d} ^{\alpha}>0.
Let y_(0)in Xy_{0} \in X such that F_(d)(f)(y_(0)) <= F_(d)(f)(x)F_{d}(f)\left(y_{0}\right) \leq F_{d}(f)(x), for all x in Xx \in X. Since F_(d)(f)(y_(0))=i n f[f(y)+||f|_(Y,d)^(alpha)d^(alpha)(y_(0),y)]F_{d}(f)\left(y_{0}\right)=\inf \left[f(y)+\|\left. f\right|_{Y, d} ^{\alpha} d^{\alpha}\left(y_{0}, y\right)\right], for every n inNn \in \mathbb{N} there exist y_(n)in Yy_{n} \in Y such that
The inequalities F_(d)(f)(y_(0)) <= F_(d)(f)(y_(n))=f(y_(n))F_{d}(f)\left(y_{0}\right) \leq F_{d}(f)\left(y_{n}\right)=f\left(y_{n}\right), imply f(y_(n))+||f|_(Y,d)^(alpha)d^(alpha)(y_(0),y_(n)) < f(y_(n))+||f|_(Y,d)^(alpha)n^(-alpha)f\left(y_{n}\right)+ \left.\left\|\left.f\right|_{Y, d} ^{\alpha} d^{\alpha}\left(y_{0}, y_{n}\right)<f\left(y_{n}\right)+\right\| f\right|_{Y, d} ^{\alpha} n^{-\alpha}, so that d(y_(0),y_(n)) < (1)/(n),n inNd\left(y_{0}, y_{n}\right)<\frac{1}{n}, n \in \mathbb{N}, i.e., d(y_(0),y_(n))rarr0d\left(y_{0}, y_{n}\right) \rightarrow 0. The sequence (y_(n))_(n >= 1)\left(y_{n}\right)_{n \geq 1} in YY is dd-convergent to y_(0)y_{0}, and since YY is dd-closed, y_(0)in Yy_{0} \in Y. 2^(0)2^{0} Let y_(0)in Yy_{0} \in Y be a global maximum point of f inLambda_(alpha)(Y)f \in \Lambda_{\alpha}(Y).
Then for every y in Yy \in Y,
If x!in Y,Yx \notin Y, Y being bar(d)\bar{d}-closed, there exists eta > 0\eta>0 such that bar(d)(x,y) > eta\bar{d}(x, y)>\eta (i.e. d(y,x) > etad(y, x)>\eta ) for all y in Yy \in Y. Therefore
{:[G_(d)(f)(x)=s u p_(y in Y){f(y)-||f|_(Y,d)^(alpha) bar(d)^(alpha)(x,y)}],[=s u p_(y in Y){f(y)-||f|_(Y,d)^(alpha)d^(alpha)(y,x)}],[ <= s u p_(y in Y){f(y)-||f|_(Y,d)^(alpha)eta^(alpha)}],[ <= f(y_(0))-||f|_(Y,d)^(alpha)eta^(alpha) <= f(y_(0))]:}\begin{aligned}
G_{d}(f)(x) & =\sup _{y \in Y}\left\{f(y)-\|\left. f\right|_{Y, d} ^{\alpha} \bar{d}^{\alpha}(x, y)\right\} \\
& =\sup _{y \in Y}\left\{f(y)-\|\left. f\right|_{Y, d} ^{\alpha} d^{\alpha}(y, x)\right\} \\
& \leq \sup _{y \in Y}\left\{f(y)-\|\left. f\right|_{Y, d} ^{\alpha} \eta^{\alpha}\right\} \\
& \leq f\left(y_{0}\right)-\|\left. f\right|_{Y, d} ^{\alpha} \eta^{\alpha} \leq f\left(y_{0}\right)
\end{aligned}
It follows that G_(d)(f)(x) <= f(y_(0))G_{d}(f)(x) \leq f\left(y_{0}\right), for all x in Xx \in X.
Conversely, suppose that y_(0)in Xy_{0} \in X is a global maximum point for G_(d)(f)G_{d}(f) in XX.
Case I ||f|_(Y,d)^(alpha)=0\|\left. f\right|_{Y, d} ^{\alpha}=0.
In this case, because f=G_(d)(f)|_(Y)f=\left.G_{d}(f)\right|_{Y} and ||G_(d)(f)|_(X,d)^(alpha)=||f|_(Y,d)^(alpha)=0\left.\left\|\left.G_{d}(f)\right|_{X, d} ^{\alpha}=\right\| f\right|_{Y, d} ^{\alpha}=0 it follows that ff and G_(d)(f)G_{d}(f) are equal with the same constant.
Case II ||f|_(Y,d)^(alpha) > 0\|\left. f\right|_{Y, d} ^{\alpha}>0.
Let y_(0)in Xy_{0} \in X such that G_(d)(f)(y_(0)) > G_(d)(f)(x),x in XG_{d}(f)\left(y_{0}\right)>G_{d}(f)(x), x \in X.
Since
G_(d)(f)(y_(0))=s u p_(y in Y)[f(y)-||f|_(Y,d)^(alpha)d^(alpha)(y,y_(0))]G_{d}(f)\left(y_{0}\right)=\sup _{y \in Y}\left[f(y)-\|\left. f\right|_{Y, d} ^{\alpha} d^{\alpha}\left(y, y_{0}\right)\right]
for every n inNn \in \mathbb{N}, there exists y_(n)in Yy_{n} \in Y such that
The inequalities G_(d)(f)(y_(0)) >= G_(d)(f)(y_(n))=f(y_(n))G_{d}(f)\left(y_{0}\right) \geq G_{d}(f)\left(y_{n}\right)=f\left(y_{n}\right), imply f(y_(n))-||f|_(Y,d)^(alpha)d^(alpha)(y_(n),y_(0)) > f(y_(n))-(1)/(n^(alpha))||f|_(Y,d)^(alpha)f\left(y_{n}\right)-\left.\left\|\left.f\right|_{Y, d} ^{\alpha} d^{\alpha}\left(y_{n}, y_{0}\right)>f\left(y_{n}\right)-\frac{1}{n^{\alpha}}\right\| f\right|_{Y, d} ^{\alpha}, so that d(y_(n),y_(0)) < (1)/(n),n inNd\left(y_{n}, y_{0}\right) <\frac{1}{n}, n \in \mathbb{N}, i.e., d(y_(n),y_(0))rarr0d\left(y_{n}, y_{0}\right) \rightarrow 0. This means that the sequence (y_(n))_(n >= 1)\left(y_{n}\right)_{n \geq 1} is bar(d)\bar{d}-convergent to y_(0)y_{0}. Since YY is bar(d)\bar{d}-closed, y_(0)in Yy_{0} \in Y.
Remark 3. By Theorem 2.6 in [14], it follows that every f inLambda_(alpha,0)(Y,d)f \in \Lambda_{\alpha, 0}(Y, d) is lower semicontinuous on ( Y,d^(alpha)Y, d^{\alpha} ) and attains its minimum on YY, provided that YY is d^(alpha)d^{\alpha}-sequentially compact. Also, every f inLambda_(alpha,0)(Y,d)f \in \Lambda_{\alpha, 0}(Y, d) is upper semicontinuous on ( Y, bar(d^(alpha))Y, \overline{d^{\alpha}} ) and attains its maximum value on YY whenever YY is bar(d^(alpha))\overline{d^{\alpha}}-sequentially compact.
If YY is (d^(alpha), bar(d^(alpha)))\left(d^{\alpha}, \overline{d^{\alpha}}\right)-sequentially compact and ( X,d^(alpha)X, d^{\alpha} ) is a T_(1)T_{1}-topological space, every f inLambda_(alpha,0)(Y,d)f \in \Lambda_{\alpha, 0}(Y, d) attains both the global minimum and the global maximum on YY. Moreover, the sequential method for the calculation of the global extremum (maximum and/or minimum) of ff, ([14], Th. 3.1) is applicable.
Acknowledgements
I would like to thank the referees for the valuable suggestions in improving the structure of the paper.
References
[1] Breckner, W.W.: Hölder continuity of certain generalized convex functions. Optimization 28, 201-209 (1994)
[2] Cobzaş, S.: Functional analysis in asymmetric normed spaces, arXiv:1006. 1175 v 1.
[3] Collins, J., Zimmer, J.: An asymmetric Arzelà-Ascoli theorem. Topol. Appl. 154, 2312-2322 (2007)
[4] Doagooei, A.R., Mohebi, H.: Optimization of the difference of ICR functions. Nonlinear Anal. Theor. Method Appl. 71, 4493-4499 (2009)
[5] Flecther, P., Lindgren, W.F.: Quasi-uniform spaces. Dekker, New York (1982)
[6] Krein, M.G., Nudel'man, A.A.: The Markov moment problem and extremum problems, Nauka, Moscow 1973 (in Russian), English translation . American Mathematical Society, Providence (1977)
[7] Künzi, H.-P.A: Nonsymmetric distances and their associated topologies: about the origin of basic ideas in the area of asymmetric topology. In: Aull, C.E., Lowen, R. Handbook of the history of general topology, pp. 853-968. Kluwer Academic Publishers, Dordrecht (2001)
[8] Martinez-Legaz, J.E., Rubinov, A.M., Schaible, S.: Increasing quasi-concave coradiant functions with applications in mathematical economics. Math. Oper. Res. 61, 261-280 (2005)
[9] Matouškova, E.: Extensions of continuous and Lipschitz functions. Canad. Math. Bull. 43(2), 208-217 (2000)
[10] McShane, E.T.: Extension of range of functions. Bull. Am. Math. Soc. 40, 837-842 (1934)
[11] Mennucci, A.: On asymmetric distances, Tech. Rep. Scuola Normale Superiore, Pisa (2004)
[12] Mustăţa, C.: Extension of semi-Lipschitz functions on quasi-metric spaces. Rev. Anal. Numér. Théor. Approx. 30, 61-67 (2001)
[13] Mustăţa, C.: A Phelps type theorem for spaces with asymmetric norms. Bull. Stiinţ. Univ. Baia Mare, Ser. B. Matematică-Informatică 18, 275-280 (2002)
[14] Mustăţa, C.: On the approximation of the global extremum of a semi-Lipschitz function. Mediterr. J. Math. 6, 169-180 (2009)
[15] Mustăţa, C.: Extensions of semi-Hölder real valued functions on a quasi-metric space. Rev. Anal. Numér. Théor. Approx. 38, 164-169 (2009)
[16] Phelps, R.R.: Uniqueness of Hahn-Banach extension and unique best approximation. Trans. Am. Math. Soc. 95, 238-255 (1960)
[17] Romaguera, S., Sanchis, M.: Semi-Lipschitz functions and best approximation in quasi-metric spaces. J. Approx. Theor. 103, 292-301 (2000)
[18] Romaguera, S., Sanchis, M.: Properties of the normed cone of semi-Lipschitz functions. Acta Math. Hungar 108, 55-70 (2005)
C. Mustăţa
"T. Popoviciu" Institute of Numerical Analysis
P.O. Box 68-1