4
$\begingroup$

I know that: Let X be a uniformly convex Banach space, $a\in X$ and $C\subset X$ closed and convex, then there is a unique $b\in C$ with $\left\Vert a-b\right\Vert=\inf_{x\in C}\left\Vert a-x \right\Vert$.

Moreover I know that: let $X$ be a Banach space, such that for every $a\in X$ and $C\subset X$ closed and convex there is at most one $b\in C$ with $\left\Vert a-b\right\Vert=\inf_{x\in C}\left\Vert a-x \right\Vert$. Then $X$ is strictly convex.

So I wonder, if the following statement is true: Let $X$ be a Banach space, such that for every $a\in X$ and $C\subset X$ closed and convex, there is a unique $b\in C$ with $\left\Vert a-b\right\Vert=\inf_{x\in C}\left\Vert a-x \right\Vert$. Then $X$ is uniformly convex.

EDIT: This statement is false, see Hsueh-Yung Lin's comment. So I should ask: Let $X$ be a Banach space, such that for every $a\in X$ and $C\subset X$ closed and convex, there is a unique $b\in C$ with $\left\Vert a-b\right\Vert=\inf_{x\in C}\left\Vert a-x \right\Vert$. Then every bounded sequence has a weakly convergent subsequence.

$\endgroup$
5
  • $\begingroup$ Well, take a finite-dimensional strictly convex (but not uniformly convex Banach space). Then compactness will ensure the existence of a point of best approximation. For an infinite-dimensional example, use a reflexive strictly (but not uniformly) convex space and argue via weak compactness. I'm not convinced this is a great question for this site: math.stackexchange.com might be better. $\endgroup$ Commented Nov 27, 2011 at 9:44
  • 1
    $\begingroup$ The following article might help you: ams.org/journals/bull/1941-47-04/S0002-9904-1941-07451-3/… $\endgroup$
    – HYL
    Commented Nov 27, 2011 at 10:27
  • 4
    $\begingroup$ @Matthew: a compactness argument actually shows that a finite dimensional strictly convex Banach space is uniformly convex. $\endgroup$ Commented Nov 27, 2011 at 12:32
  • $\begingroup$ @Hsueh-Yung Lin: interesting paper. M. Day provides a example for a reflexive strictly convex space, which is not isomorphic to a uniformly convex space. So we get the unique best approximation, because every bounded sequence admits a weakly-convergent subsequence, so I have to modify my question. Is it true, that, if every closed convex set admits a best approximation, then every bounded sequence admits a weakly-convergent subsequence. $\endgroup$ Commented Nov 27, 2011 at 17:23
  • 2
    $\begingroup$ @Mark: Ah, yes! Very silly... $\endgroup$ Commented Nov 27, 2011 at 18:40

1 Answer 1

8
$\begingroup$

Your modified question has an affirmative answer. An equivalent form, in view of the Eberlein-Smulian theorem, is whether the Banach space $X$ must be reflexive if every closed bounded non empty set admits best approximations. If $X$ is not reflexive, then by R. C. James' famous characterization of reflexivity, there is a norm one linear functional $F$ on $X$ s.t. $F$ does not achieve its norm on the closed unit ball $B_X$. Let $C:= [F=1]\cap 2B_X$. This closed bounded non empty convex set contains no point of minimal norm.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .