On the average box dimensions of graphs of typical continuous functions (Q1714965): Difference between revisions

From MaRDI portal
Set OpenAlex properties.
Created claim: Wikidata QID (P12): Q129139855, #quickstatements; #temporary_batch_1726256605485
 
Property / Wikidata QID
 
Property / Wikidata QID: Q129139855 / rank
 
Normal rank

Latest revision as of 20:43, 13 September 2024

scientific article
Language Label Description Also known as
English
On the average box dimensions of graphs of typical continuous functions
scientific article

    Statements

    On the average box dimensions of graphs of typical continuous functions (English)
    0 references
    0 references
    1 February 2019
    0 references
    For a bounded set \(E\subseteq \mathbb{R}^m\) and \(\delta >0\) let \(N_\delta(E)\) be the number of \(\delta\)-mesh cubes that intersect \(E\) and \(\Delta(E,\delta)=\frac{\log N_\delta(E)}{-\log\delta}\). The lower and upper box dimensions are defined by \(\underline{\dim}_B (E)=\liminf_{\delta\searrow 0}\Delta(E,\delta)\) and \(\overline{\dim}_B (E)=\limsup_{\delta\searrow 0}\Delta(E,\delta)\). Let \(X\) be a bounded subset of \(\mathbb{R}^d\) and \(C_u(X)\) the Banach space of all uniformly continuous real-valued functions on \(X\) equipped with the uniform norm \(\|\;\|_\infty\). It is known that \(\underline{\dim}_B(X)\leq \underline{\dim}_B(\text{graph}(f))\leq \overline{\dim}_B(\text{graph}(f))\leq \overline{\dim}_B(X)+1\) for \(f\in C_u(X)\) and \(\underline{\dim}_B(\text{graph}(f))=\underline{\dim}_B(X)\) for a typical function \(f\in C_u(X)\); moreover, if \(X\) has no isolated points, then \(\overline{\dim}_B(\text{graph}(f))=\overline{\dim}_B(X)+1\) for a typical function \(f\in C_u(X)\). Here the authors say that a typical function \(f\in C_u(X)\) has the property \(P\) if \(\{f\in C_u(X): f \text{ has property } P\}\) is co-meager. The aim of the paper is to show: ``not only is the box counting function \(\Lambda_f(\delta):=\Delta(\text{graph}(f),\delta)\) divergent as \(\delta\searrow 0\), but it is so irregular that it remains spectacularly divergent as \(\delta\searrow 0\) even after being ``averaged''. The authors illustrate their very general results by the following particular case: Define \(\Lambda_f^n(t)\) inductively by \(\Lambda_f^0(t)=\Lambda_f(e^{-t})\), \(\Lambda_f^n(t)=\frac{1}{t}\int_1^t\Lambda_f^{n-1}(s)ds\). Then a typical continuous function \(f:[0,1]^d\rightarrow\mathbb{R}\) satisfies \(\liminf_{t\rightarrow\infty}\Lambda_f^n(t)=d\), \(\limsup_{t\rightarrow\infty}\Lambda_f^n(t)=d+1\) for all \(n\in\mathbb{N}\cup \{0\}\).
    0 references
    0 references
    box dimension
    0 references
    continuous function
    0 references
    Hölder mean
    0 references
    Cesáro mean
    0 references
    Riesz-Hardy mean
    0 references
    Baire category
    0 references

    Identifiers