Symmetric seminorms and the Leibniz property (Q2400680): Difference between revisions

From MaRDI portal
Import240304020342 (talk | contribs)
Set profile property.
Set OpenAlex properties.
Property / OpenAlex ID
 
Property / OpenAlex ID: W2964300080 / rank
 
Normal rank

Revision as of 19:43, 19 March 2024

scientific article
Language Label Description Also known as
English
Symmetric seminorms and the Leibniz property
scientific article

    Statements

    Symmetric seminorms and the Leibniz property (English)
    0 references
    0 references
    29 August 2017
    0 references
    The main result of the paper is the following inequality, previously known for \(p = 2\): for every probability space \((S, {\mathcal F}, \mu)\) and \(1 \leq p < \infty\), one has: \[ \| f g - {\mathbb E} (f g) \|_p \leq \| g \|_\infty \| f - {\mathbb E} f \|_p + \| f \|_\infty \| g - {\mathbb E} g \|_p \] for all real \(f, g \in L^\infty (S, \mu)\). For that, the author proves that, if \(\| \, . \, \|\) is any symmetric norm on \({\mathbb R}^n\), then: \[ \| f g - {\mathbb E} (f g) \| \leq \| g \|_\infty \| f - {\mathbb E} f \| + \| f \|_\infty \| g - {\mathbb E} g \| \eqno (1) \] for all \(f, g \in {\mathbb R}^n\), where \({\mathbb E} f =\big[ {1 \over n} \, (f_1 + \cdots + f_n) \big] \, (1, \dots, 1)\) if \(f = (f_1, \dots, f_n) \in {\mathbb R}^n\), and uses Proposition~2.1 of \textit{Á. Besenyei} and \textit{Z. Léka} [J. Math. Anal. Appl. 429, No. 2, 1178--1189 (2015; Zbl 1335.46056)]. Two proofs of \((1)\) are given; the first one uses the Ky Fan \(k\)-norms defined by \(\| x \|_{(k)} = |x|_1^\ast + \cdots + |x|_k^\ast\), where \((|x|_1^\ast, \dots, |x|_n^\ast)\) is the non-increasing rearrangement of \(|x| = (|x_1|, \dots, |x_n|)\); the second one uses a suitable derivation \(\partial : \ell_\infty^n \to M_n ({\mathbb R}^n)\). If \(({\mathcal A}, \| \, . \, \|)\) is a unital normed algebra, a seminorm \(L\) on \({\mathcal A}\) is said to be Leibniz if \(L (1_{\mathcal A}) = 0\) and \[ L (a b) \leq \| a \| \, L (b) + \| b \| \, L (a) \] for all \(a, b \in {\mathcal A}\). It is strongly Leibniz if, moreover, \(L (a^{- 1}) \leq \| a^{- 1}\|^2 L (a)\) for all invertible elements \(a\). \textit{M. A. Rieffel} [New York J. Math. 20, 35--56 (2014; Zbl 1310.46056)] showed that, if \((S, \mu)\) a probability space, then \(L (f) = \| f - {\mathbb E} (f) \|_2\) defines a strong Leibniz semi-norm on \({\mathcal A} = L^\infty (S, \mu)\); and if \(\Omega\) is a normed bimodule over \({\mathcal A}\) and \(\partial : {\mathcal A} \to \Omega\) is a derivation satisfying \(\partial (ab) = \partial (a) b + a \partial (b)\), then \(L (a) = \| \partial a \|_\Omega\) is a strong Leibniz seminorm on \({\mathcal A}\). The author shows that such a situation arises from the Cipriani-Sauvageot differential calculus and links it to Rieffel's non-commutative Riemann metric. An example of a Leibniz seminorm on \({\mathbb R}^3\) which is not strongly Leibniz is given.
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    central moments
    0 references
    derivation
    0 references
    Leibniz inequality
    0 references
    standard deviation
    0 references
    symmetric norm
    0 references
    0 references