On the representation of smooth functions on the sphere using finitely many bits (Q1780689): Difference between revisions

From MaRDI portal
RedirectionBot (talk | contribs)
Removed claims
RedirectionBot (talk | contribs)
Changed an Item
Property / author
 
Property / author: Hrushikesh N. Mhaskar / rank
 
Normal rank
Property / reviewed by
 
Property / reviewed by: Yu. I. Makovoz / rank
 
Normal rank

Revision as of 16:01, 19 February 2024

scientific article
Language Label Description Also known as
English
On the representation of smooth functions on the sphere using finitely many bits
scientific article

    Statements

    On the representation of smooth functions on the sphere using finitely many bits (English)
    0 references
    13 June 2005
    0 references
    The paper deals with approximation of smooth functions defined on the \(q\)-dimensional Euclidean sphere \(S^q\) by polynomial operators. Let \({\mathcal P}_l(t)\) be the Legendre polynomial of degree \(l\) and let \(h(t)\) be some \([0, \infty) \to {\mathbb R}\) function vanishing for sufficiently large \(t\). For \(t \in {\mathbb R}\) and \(n=0,1,\ldots\), let \[ \Phi_n(t)=\sum_{l=0}^\infty h(l/2^n) {d_l^q \over \omega_q}{\mathcal P}_l(t), \] where \(d_l^q\) is the linear dimension of the set of all spherical harmonics of degree \(l\) and \(\omega_q\) is the volume of \(S^q\). Given a sequence \(\{\mu_n\}\) of signed Borel measures on \(S^q\), the summability operator is defined by \[ \sigma_n(f, {\mathbf x})=\int_{S_q} \Phi_n ({\mathbf x}\cdot \xi)\,d\mu_n(\xi). \] The difference \(\tau_n(f)=\sigma_n(f)-\sigma_{n-1}(f)\) is the polynomial frame operator. The most interesting cases are when \(\mu_n\) are the volume measures on \(S^q\) for all \(n\) and when \(\mu_n\) are the so-called M-Z quadrature measures. In the latter case, the author proves that under certain uniformity conditions \(f=\sum_{n=0}^\infty \tau_n(f)\), with the series convergent in the sense of \(L_p(S^q)\). Various properties of this expansion are proved. In particular, by uniform quantization of the values of polynomial operators one can obtain a bit coding of functions \(f\) from the Besov classes, with the number of bits that is within only a logarithmic factor of their metric entropy. Moreover, the coding has a localization property: from the total set of bits one can identify for each spherical cap a certain subset of bits that represents the function on this cap. The cardinality of this subset depends on the local smoothness of the function and the volume of the cap.
    0 references
    approximation on the sphere
    0 references
    polynomial frames
    0 references
    metric entropy
    0 references

    Identifiers

    0 references
    0 references
    0 references
    0 references
    0 references
    0 references