--- title: Restricted Center Map in Banach Spaces url: https://www.emergentmind.com/topics/restricted-center-map type: topic --- # Restricted Center Map in Banach Spaces Searching arXiv for the main paper and closely related work on restricted center maps, restricted Chebyshev centers, and center-as-equivariant-map formulations. I’m checking arXiv records for the principal Banach-space treatment and related formulations of “restricted center map” to ground the article in current literature. In Banach-space approximation theory, the restricted center map is the set-valued assignment that sends a bounded closed set \(F\) to its set of best Chebyshev centers constrained to lie in a fixed closed subset \(V\) of a real Banach space \(X\). In the formulation developed in "On stability of restricted center properties and continuity of restricted center map under \(\ell_p\)-direct sum" [2510.01955], the central objects are the restricted radius \[ \operatorname{rad}_V(F)=\inf\{r(v,F):v\in V\}, \qquad r(x,F)=\sup\{\|x-a\|:a\in F\}, \] and the restricted center set \[ Z_V(F)=\{v\in V:r(v,F)=\operatorname{rad}_V(F)\}. \] The map \(F\mapsto Z_V(F)\) generalizes the ordinary Chebyshev center map by imposing a geometric constraint on admissible centers, and its modern theory is organized around existence, Hausdorff continuity, and stability under direct-sum constructions. ## 1. Basic setting and formal definition The theory is formulated in real Banach spaces \(X\), with \(B_X\) and \(S_X\) denoting the closed unit ball and unit sphere. For a nonempty subset \(M\subset X\), the standard hyperspaces are \(\operatorname{CL}(M)\) for nonempty closed subsets, \(\operatorname{CB}(M)\) for nonempty closed bounded subsets, \(\operatorname{CC}(M)\) for nonempty closed convex subsets, and \(\mathcal K(M)\) for nonempty compact subsets. Fix \(V\in \operatorname{CL}(X)\). For \(F\in \operatorname{CB}(X)\), the restricted center set is \[ Z_V(F)=\{v\in V:r(v,F)=\operatorname{rad}_V(F)\}, \] where \[ \operatorname{rad}_V(F)=\inf\{r(v,F):v\in V\}. \] The pair \((V,F)\) has the restricted center property, abbreviated r.c.p., if \(Z_V(F)\neq\emptyset\). If \(\mathcal F\subset \operatorname{CB}(X)\), then \((V,\mathcal F)\) has r.c.p. when every \(F\in\mathcal F\) has a restricted center. The relevant hyperspace topology is induced by the Hausdorff metric \[ H(A,B)=\inf\{r>0:A\subset B+rB_X,\ \ B\subset A+rB_X\}. \] With this metric, the restricted center map is \[ Z_V:(\mathcal F,H)\longrightarrow (\operatorname{CB}(V),H),\qquad F\mapsto Z_V(F). \] A standard enlargement is \[ Z_V(F,\delta)=\{v\in V:r(v,F)\le \operatorname{rad}_V(F)+\delta\}, \] so that \(Z_V(F,0)=Z_V(F)\). This approximate-center set is the basic device for formulating stability and continuity. The restriction may be imposed by a subspace, but the formalism is broader: \(V\) is only assumed closed. This makes the framework applicable simultaneously to subspaces, balls, and other closed constraint sets. ## 2. Stability properties and continuity criteria The modern theory distinguishes three stability properties for the pair \((V,\mathcal F)\) [2510.01955]. Property-\((P_1)\) requires r.c.p. and, for every \(\varepsilon>0\), existence of \(\delta>0\) such that \[ Z_V(F,\delta)\subset Z_V(F)+\varepsilon B_X \] for each fixed \(F\in\mathcal F\). Property-\((P_2)\) has the same inclusion, but with a single \(\delta\) working uniformly for every \(G\in\mathcal F\). Property-\((lP_2)\) interpolates between them: for every \(\varepsilon>0\) and \(F\in\mathcal F\), there exists \(\delta>0\) such that \[ Z_V(G,\delta)\subset Z_V(G)+\varepsilon B_X \quad\text{whenever }G\in\mathcal F\text{ and }H(F,G)<\delta. \] | Property | Quantifier pattern | Continuity consequence | |---|---|---| | \((P_1)\) | \(\delta\) may depend on the fixed set \(F\) | \(Z_V\) is uHsc | | \((lP_2)\) | \(\delta\) may depend on the base set \(F\), but works for all nearby \(G\) | equivalent to \((P_1)\) + lHsc; hence Hausdorff continuous | | \((P_2)\) | one \(\delta\) works uniformly for all \(G\in\mathcal F\) | uniformly Hausdorff continuous | The sequential characterizations clarify the distinction. Property-\((P_1)\) is equivalent to r.c.p. plus the requirement that every minimizing sequence for a fixed \(F\) can be asymptotically approximated by a sequence of exact restricted centers of that same \(F\). Property-\((P_2)\) is equivalent to r.c.p. plus the analogous statement for varying sets \(F_n\), with \(y_n\in Z_V(F_n)\) and \(\|v_n-y_n\|\to 0\). In the language of the paper, \((P_1)\) is local stability for one set, whereas \((P_2)\) is uniform stability over the whole class. For a set-valued map \(T:(N,d)\to(\operatorname{CB}(X),H)\), upper Hausdorff semi-continuity at \(a_0\) means \[ T(a)\subset T(a_0)+\varepsilon B_X \] for all \(a\) sufficiently close to \(a_0\); lower Hausdorff semi-continuity means \[ T(a_0)\subset T(a)+\varepsilon B_X \] for all nearby \(a\). Hausdorff continuity is the conjunction of both, and uniform Hausdorff continuity requires a \(\delta\) independent of the base points. If all values are singletons, uHsc, lHsc, and Hausdorff continuity coincide. The key structural theorem identifies \((lP_2)\) exactly with the missing lower-semicontinuity ingredient: \[ (P_2)\Rightarrow (lP_2)\Rightarrow (P_1), \] and \[ (lP_2)\iff (P_1)\text{ and }Z_V\text{ is lHsc}. \] Hence \((P_1)\) yields upper Hausdorff semi-continuity, \((lP_2)\) yields full Hausdorff continuity, and \((P_2)\) yields uniform Hausdorff continuity. ## 3. Behavior under \(\ell_p\)-direct sums A major structural result is that restricted center phenomena transfer through \(\ell_p\)-direct sums [2510.01955]. For \(10\) there exists \(j\) such that \[ \sum_{i>j}\|y_{n,i}\|^p<\varepsilon^p \quad\text{for all }n. \] Using this, the paper proves \[ (Y_i,F_i)\text{ has }(P_1)\text{ for every }i \iff (Y,F)\text{ has }(P_1). \] So property-\((P_1)\) is stable under infinite \(\ell_p\)-direct sums. Property-\((P_2)\) is more delicate. For finite index sets \(I\), one has \[ (Y_i,\mathcal F_i)\text{ has }(P_2)\text{ for all }i \iff (Y,\mathcal F)\text{ has }(P_2), \] for the normalized families with restricted radius \(1\). By contrast, Example 3.6 constructs an infinite \(\ell_p\)-sum in which each coordinate pair has \((P_2)\), yet the sum fails \((P_2)\): there exist \(y_n\in Y\) and \(G_n\in\mathcal F\) with \[ r(y_n,G_n)-\operatorname{rad}_Y(G_n)\to 0 \] but \[ \operatorname{dist}(y_n,Z_Y(G_n))=1 \quad\forall n. \] This is the basic obstruction separating finite and infinite direct sums. The paper’s new property-\((lP_2)\) restores stability at the continuity level. Under the standing r.c.p. hypotheses, \[ Z_{Y_i}\text{ lHsc on }\operatorname{CB}(X_i)\ \forall i \iff Z_Y\text{ lHsc on }\mathcal P(X), \] and similarly for uHsc. Combining these equivalences with the characterization of \((lP_2)\) yields \[ (Y_i,\operatorname{CB}(X_i))\text{ has }(lP_2)\ \forall i \iff (Y,\mathcal P(X))\text{ has }(lP_2). \] A direct application is the metric-projection statement that \[ Y_i\text{ locally uniformly strongly proximinal on }X_i\ \forall i \iff Y\text{ locally uniformly strongly proximinal on }X. \] ## 4. Restricted Chebyshev-center maps and earlier Banach-space developments Earlier work placed the restricted center map in the broader setting of restricted Chebyshev centers. The 2021 paper "Some stability properties of restricted Chebyshev centers in Banach spaces" studies existence of Chebyshev centers of closed bounded subsets of \(C(K,X)\) for an extremally disconnected compact Hausdorff space \(K\) and a finite dimensional Banach space \(X\), existence of Chebyshev centers of compact subsets of \(M\)-ideals in \(C(K,X)\), stability of existence of restricted Chebyshev centers in spaces of vector-valued bounded functions, and the dependence of continuity properties of the Chebyshev-center map of \(X\) on those of \(C(K,X)\) [2108.00629]. A more explicit restricted-center-map formalism appears in "On property-\((P_1)\) and semi-continuity properties of restricted Chebyshev-center maps in \(\ell_\infty\)-direct sums" [2303.10676]. There the notation is \[ \operatorname{rad}_V(B)=\inf_{v\in V}\sup_{b\in B}\|v-b\|, \qquad \operatorname{cent}_V(B)=\{v\in V:r(v,B)=\operatorname{rad}_V(B)\}, \] for a nonempty closed convex set \(V\subset X\) and \(B\in CB(X)\). The resulting restricted Chebyshev-center map \[ \operatorname{cent}_V(\cdot):\mathcal F\to CB(X) \] is the same object up to notation: admissible centers are constrained to lie in \(V\), while the set being centered lives in the ambient space \(X\). In that formulation, property-\((P_1)\) is presented as a set-valued generalization of strong proximinality. When \(\mathcal F\) consists of singletons, r.c.p. becomes proximinality and \((P_1)\) becomes strong proximinality. Mach’s theorem, as quoted there, states that property-\((P_1)\) implies upper Hausdorff semi-continuity of the restricted center map. The paper then proves stability of \((P_1)\), r.c.p., and semi-continuity under \(\ell_\infty\)-direct sums, and derives a concrete continuity theorem: if \(Y\) is a proximinal finite codimensional subspace of \(c_0\), then the closed unit ball \(B_Y\) satisfies property-\((P_1)\) for the non-empty closed bounded subsets of \(\ell_\infty\), and the restricted Chebyshev-center map \[ \operatorname{cent}_{B_Y}(\cdot):CB(\ell_\infty)\to CB(\ell_\infty) \] is Hausdorff metric continuous. This earlier \(\ell_\infty\)-sum theory and the later \(\ell_p\)-sum theory address the same general problem—stability of constrained-center assignments under Banach-space constructions—but with different ambient geometries and different continuity thresholds. ## 5. Geometric characterizations, finite-dimensional effects, and sharpness The restricted center map is tightly connected with rotundity properties of the ambient space relative to the constraining subspace \(Y\) [2510.01955]. The paper uses quasi uniform rotundity with respect to \(Y\), defined by the condition that for every \(\varepsilon>0\) there exists \(\delta>0\) such that for every \(v\in Y\) there is \(w\in Y\) with \(\|w\|\le \varepsilon\) and \[ B[0,1]\cap B[v,1-\delta]\subset B[w,1-\delta]. \] For the bounded-radius family \[ \mathcal F=\{F\in\operatorname{CB}(X):\operatorname{rad}_Y(F)\le a\}, \] the following are equivalent: \(X\) is quasi uniformly rotund with respect to \(Y\); \((Y,\mathcal F)\) has property-\((P_2)\); and the restricted center map \(Z_Y\) is nonempty-valued and uniformly Hausdorff continuous on \((\mathcal F,H)\). Property-\((P_2)\) is therefore not merely a hyperspace condition; on bounded-radius classes it is equivalent to a geometric rotundity requirement. Uniform rotundity with respect to a subspace gives a sharper criterion in the presence of uniqueness. The definition requires that whenever \((x_n)\) and \((y_n)\) lie in \(S_X\), satisfy \(x_n-y_n\in Y\), and obey \(\|x_n+y_n\|\to 1\), then \(\|x_n-y_n\|\to 0\). For \[ \mathcal F_1=\{F\in\operatorname{CB}(X):\operatorname{rad}_Y(F)\le a\}, \qquad \mathcal F_2=\{F\in\mathcal K(X):\operatorname{rad}_Y(F)\le a\}, \] uniform rotundity with respect to \(Y\) is equivalent to property-\((P_2)\) together with singleton-valued restricted centers on \(\mathcal F_1\), and equivalently on \(\mathcal F_2\). The sharpness theory is equally central. The paper gives examples showing that the three properties are genuinely distinct: there are spaces with \((P_1)\) but not \((lP_2)\), and spaces with \((lP_2)\) but not \((P_2)\). Example 3.12(1) uses a normed \(\mathbb R^3\) where the metric projection onto a one-dimensional subspace is not lHsc, so \((lP_2)\) fails although Mach’s result gives \((P_1)\). Example 3.12(2) produces a rotund but not URED space with \((P_1)\) and \((lP_2)\) but not \((P_2)\) on certain classes. Example 3.14 shows the same separations can persist even for hyperplanes. Two mechanisms collapse these distinctions. First, compactness of the family \(\mathcal F\) in Hausdorff metric yields uniformization: if \(\mathcal F\) is compact, then \((lP_2)\) and \((P_2)\) coincide. Consequently, in finite-dimensional \(X\), for \[ \mathcal F=\{F\in\operatorname{CB}(X):\operatorname{rad}_Y(F)=1\}, \] one has \[ (Y,\operatorname{CB}(X))\text{ has }(lP_2) \iff (Y,\mathcal F)\text{ has }(P_2). \] Second, URED geometry identifies \((P_1)\) and \((lP_2)\) for convex constraint sets: if \(X\) is URED and \(C\in\operatorname{CC}(X)\), then \((C,\mathcal F)\) has \((P_1)\) if and only if it has \((lP_2)\). For finite-dimensional subspaces \(Y\subset X\), URED also yields \((P_2)\) on the normalized family \(\{F:\operatorname{rad}_Y(F)=1\}\). ## 6. Terminological variants in geometric-object theory Outside Banach-space approximation theory, the phrase “center map” is often used in a different sense. In "The concept of center as an equivariant map and a proof of an analogue of the center conjecture for equifacetal simplices" [2301.09945], a center is a \(G\)-equivariant map \[ \mathfrak Z:A\to X \] between \(G\)-spaces, typically with \(A\) a family of subsets of \(X\) carrying the pointwise action. The main existence theorem states that, under the fixed-point compatibility condition \[ A^g\neq\varnothing \Rightarrow X^g\neq\varnothing, \] for every \(V\in A\) and every point \(P\in X\) fixed by the symmetry group \(G_V\), there exists an equivariant map \(\mathfrak Z\) such that \(\mathfrak Z(V)=P\). In that literature, the restriction is not a Banach-space constraint \(V\subset X\) on admissible centers; rather, it is a symmetry restriction on allowable values at a given object. The review "On the concept of center for geometric objects and related problems" gives the same broad viewpoint for the similarity group \(S(n)\) [2501.11444]. There a center is an equivariant map \[ \mathfrak X:\mathcal O\to \mathbb R^n, \qquad T(\mathfrak X(O))=\mathfrak X(T(O)), \] defined on a space of geometric objects such as multisets, polygons, or Borel sets. The paper repeatedly treats restricted domains—nondegenerate triangles \(\mathcal P_3'\), simple polygons \(\mathcal P_m'\), multisets with unique medoid \(\mathcal M_m'\), or Borel sets with \(0