Almost minimal orthogonal projections

DOI10.1007/S11856-021-2163-8MaRDI QIDQ2048865zbMATH OpenOpenAlexFDO

Publication date 24 August 2021

Published in Israel Journal of Mathematics (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/2001.08698

finite-dimensional Banach space projection

Norms of matrices, numerical range, applications of functional analysis to matrix theory (15A60) Isometric theory of Banach spaces (46B04) Geometry and structure of normed linear spaces (46B20) Linear operators belonging to operator ideals (nuclear, (p)-summing, in the Schatten-von Neumann classes, etc.) (47B10) Convexity and finite-dimensional Banach spaces (including special norms, zonoids, etc.) (aspects of convex geometry) (52A21)

Abstract: The projection constant

P i (E) := P i (E, e l l_{i} n f t y)

of a finite-dimensional Banach space

E s u b s e t e l l_{i} n f t y

is by definition the smallest norm of a linear projection of

e l l_{i} n f t y

onto

E

. Fix

n g e q 1

and denote by

P i_{n}

the maximal value of

P i (c d o t)

amongst

n

-dimensional real Banach spaces. We prove for every

v a r e p s i l o n > 0

that there exist an integer

d g e q 1

and an

n

-dimensional subspace

E s u b s e t e l l_{1}^{d}

such that

P i_{n} l e q P i (E, e l l_{1}^{d}) + 2 v a r e p s i l o n

and the orthogonal projection

P c o l o n e l l_{1}^{d} o E

is almost minimal in the sense that

l V e r t P V e r t l e q P i (E, e l l_{1}^{d}) + v a r e p s i l o n

. As a consequence of our main result, we obtain a formula relating

P i_{n}

to smallest absolute value row-sums of orthogonal projection matrices of rank

n

.

Review text

This paper makes a significant contribution to the problem of finding the maximum projection constant for the class of Banach spaces of a given finite dimension. More precisely, for any integer \(n\) and \(\varepsilon>0\), it shows that there is an integer \(d>n\) and an \(n\)-dimensional subspace of \(\mathbb{R}^d\) (equipped with the \(\ell_1\) norm) whose relative projection constant is almost maximal (i.e., within \(\varepsilon\) of the maximum). Moreover, the orthogonal projection (with respect to a suitable orthonormal basis of \(\mathbb{R}^d\)) onto this subspace has almost minimal norm.

Recommendations

Cites work

Cited in

(12)

This page was built for publication: Almost minimal orthogonal projections

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2048865)