Average Nikolskii factors for random diffusion polynomials on closed Riemannian manifolds
Yun Ling, Heping Wang · 2025
Read Average Nikolskii factors for random diffusion polynomials on closed Riemannian manifolds by Yun Ling, Heping Wang (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-06505.txt.
Introduction
Yun Ling and Heping Wang compute typical size ratios for random polynomials on curved spaces. Their paper appeared in two thousand twenty-five. A Nikolskii factor compares one norm of a polynomial against another. Worst-case bounds grow fast with degree. Random choices behave far better, usually. The setting is diffusion polynomials on closed Riemannian manifolds. Coefficients are independent Gaussian noise.
The authors find the exact order of the average factor. For finite target norms, the average stays constant as degree grows. For the supremum norm, it grows only like a root of a logarithm. Both crush the worst-case rates. The proof works through eigenpairs of the Laplace-Beltrami operator. Approximation theory gains sharp expectations for random expansions.
Catalog record
| CREATOR | Yun Ling, Heping Wang |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 38 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-06505
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Average Nikolskii factors for random diffusion polynomials on closed Riemannian manifolds Yun Ling, Heping Wang Abstract . For 1 ≤ p , q ≤ ∞ 1\leq p,q\leq\infty , the Nikolskii factor for a diffusion polynomial P 𝐚 P_{{\mathbf{a}}} of degree at most n n is defined by N p , q ( P 𝐚 ) = ‖ P 𝐚 ‖ q ‖ P 𝐚 ‖ p , P 𝐚 ( 𝐱 ) = ∑ k : λ k ≤ n a k ϕ k ( 𝐱 ) , N_{p,q}(P_{{\mathbf{a}}})=\frac{\|P_{{\mathbf{a}}}\|_{q}}{\|P_{{\mathbf{a}}}\|_{p}},\ \ P_{{\mathbf{a}}}({\bf x})=\sum_{k:\lambda_{k}\leq n}a_{k}\phi_{k}({\bf x}), where 𝐚 = { a k } λ k ≤ n {\mathbf{a}}=\{a_{k}\}_{{\lambda}_{k}\leq n} , and { ( ϕ k , − λ k 2 ) } k = 0 ∞ \{(\phi_{k},-\lambda_{k}^{2})\}_{k=0}^{\infty} are the eigenpairs of the Laplace-Beltrami operator Δ 𝕄 \Delta_{\mathbb{M}} on a closed smooth Riemannian manifold 𝕄 \mathbb{M} with normalized Riemannian measure. We study this average Nikolskii factor for random diffusion polynomials with independent N ( 0 , σ 2 ) N(0,\sigma^{2}) coefficients and obtain the exact orders. For 1 ≤ p < q < ∞ 1\leq p<q<\infty , the average Nikolskii factor is of order n 0 n^{0} (i.e., constant), as compared to the worst case bound of order n d ( 1 / p − 1 / q ) n^{d(1/p-1/q)} , and for 1 ≤ p < q = ∞ 1\leq p<q=\infty , the average Nikolskii factor is of order ( ln n ) 1 / 2 (\ln n)^{1/2} as compared to the worst case bound of order n d / p n^{d/p} . 2020 Mathematics Subject Classification 26D05, 42A05 1. Introduction Let {M} be a closed connected smooth d -dimensional Riemannian manifold equipped with the normalized Riemannian measure \mu . Here a closed manifold is a compact manifold without boundary. For 1\leq p<\infty , let L_{p}({M}) be the set of all complex valued measurable function defined on {M} endowed with finite norm \|f\|_{p}:=\left(\int_{{M}}|f({\bf x})|^{p}{\rm d}\mu({\bf x})\right)^{1/p}, and let L_{\infty}({M})\equiv C({M}) be the set of all complex valued continuous functions defined on {M} endowed with uniform norm \|f\|_{\infty}:=\sup\limits_{{\bf x}\in{M}}|f({\bf x})|. In particular, L_{2}({M}) is a Hilbert space with inner product (1.1) \langle f,g\rangle:=\int_{{M}}f({\bf x})\overline{g({\bf x})}{\rm d}\mu({\bf x}),\ {\rm for}\ f,g\in L_{2}({M}), where \overline{g} is the complex conjugate to g . Let \Delta_{{M}} be the Laplace-Beltrami operator on {M} , which is self-adjoint in L_{2}({M}) . This operator has a sequence of eigenvalues \{-\lambda_{k}^{2}\}_{k=0}^{\infty} arranged by 0=\lambda_{0}\leq\lambda_{1}\leq\dots\leq\lambda_{k}\leq\dots\to\infty, and a complete orthonormal system of eigenfunctions \{\phi_{k}\}_{k=0}^{\infty} . Without loss of generality, we choose each \phi_{k} to be real-valued and \phi_{0}=1 . Then \{\phi_{k}\}^{\infty}_{k=0} is an orthonormal basis for L_{2}(\mathbb{M}) . For each n=0,1,2,\dots , with \mathcal{P}_{n} we denote the set of “polynomials” of degree at most n on {M} by \mathcal{P}_{n}:={\rm span}\left\{\phi_{k}:\,\lambda_{k}\leq n\right\}. This space is called the diffusion polynomial space of degree n on {M} , and an element of \mathcal{P}_{n} is called a diffusion polynomial of degree at most n . We have two-sides Wely’s law with exponent d and constants c_{2}(d)>c_{1}(d)>0 (see [ 18 , 25 , 26 , 39 ] ), i.e., (1.2) c_{1}(d)n^{d}\leq\sum_{k:\lambda_{k}\leq n}|\phi_{k}({\bf x})|^{2}\leq c_{2}(d)n^{d},\ \ {\bf x}\in{M}, which, by integrating on {M} with respect to {\bf x} , implies that the dimension N of the space \mathcal{P}_{n} satisfies (1.3) N:=\dim\mathcal{P}_{n}:=\#\{\phi_{k}:\lambda_{k}\leq n\}\asymp n^{d}, where \#E denotes the number of element in the set E . Here we use the notation A_{n}\asymp B_{n} to express A_{n}\ll B_{n} and A_{n}\gg B_{n} , and A_{n}\ll B_{n} ( A_{n}\gg B_{n} ) means that there exists a constant c>0 independent of n such that A_{n}\leq cB_{n} ( A_{n}\geq cB_{n} ). Let P be a diffusion polynomial of degree at most n . Then, P is of the form P_{{\mathbf{a}}} (1.4) P({\bf x})=P_{{\mathbf{a}}}({\bf x})=\sum_{k:\lambda_{k}\leq n}a_{k}\phi_{k}({\bf x}),\ \ {\bf x}\in\mathbb{M}, where {\mathbf{a}}=\{a_{k}\}_{\lambda_{k}\leq n}\in{R}^{N} . The classical Nikolskii inequality for diffusion polynomials (see [ 32 , Lemma 5.5] ) states that there exists a positive constant C(d) depending only on d such that for any 1\leq p,q\leq\infty and P_{{\mathbf{a}}}\in\mathcal{P}_{n} , (1.5) \|P_{{\mathbf{a}}}\|_{q}\leq C(d)n^{d(1/p-1/q)_{+}}\|P_{{\mathbf{a}}}\|_{p}, where a_{+}=\max(a,0) for a real number a . The order d(1/p-1/q)_{+} in the above inequality ( 1.5 ) is sharp. For given 1\leq p,q\leq\infty , we introduce the Nikolskii factor for P_{\mathbf{a}} by {N}_{p,q}(P_{{\mathbf{a}}}):=\frac{\|P_{{\mathbf{a}}}\|_{q}}{\|P_{{\mathbf{a}}}\|_{p}}. The worst Nikolskii factor for \mathcal{P}_{n} is defined by N^{\rm wor}_{p,q}(\mathcal{P}_{n}):=\sup_{0\neq P_{{\mathbf{a}}}\in\mathcal{P}_{n}}{N}_{p,q}(P_{{\mathbf{a}}})=\sup_{0\neq P_{{\mathbf{a}}}\in\mathcal{P}_{n}}\frac{\|P_{{\mathbf{a}}}\|_{q}}{\|P_{{\mathbf{a}}}\|_{p}}, where the supremum is taken over all diffusion polynomials of degree \leq n whose coefficients are not all zero. Then the above Nikolskii inequality ( 1.5 ) states (1.6) N^{\rm wor}_{p,q}(\mathcal{P}_{n})\asymp n^{d(1/p-1/q)_{+}}, i.e., that the worst case bound for {N}_{p,q}(P_{{\mathbf{a}}}) is \Theta(n^{d(1/p-1/q)_{+}}) . Now we introduce average Nikolskii factors on the compact manifold {M} . Average Nikolskii factors were first introduced and studied in [ 30 ] . Average Nikolskii factors on the torus and the sphere were investigated in [ 30 ] and [ 31 ] , respectively. A random diffusion polynomial P_{{\mathbf{a}}}\in\mathcal{P}_{n} is defined by (1.7) P_{{\mathbf{a}}}({\bf x})=\sum\limits_{k:\lambda_{k}\leq n}a_{k}\phi_{k}({\bf x}), where {\mathbf{a}}=\{a_{k}\}_{\lambda_{k}\leq n}\in\mathbb{R}^{N} , the coefficients a_{k} are independent N(0,\sigma^{2}) random variables with the common normal density functions \frac{1}{\sqrt{2\pi}{\sigma}}e^{-\frac{a_{k}^{2}}{2{\sigma}^{2}}}. That is, {\mathbf{a}}\sim N(0,{\sigma}^{2}I_{N}) is an \mathbb{R}^{N} -valued Gaussian random vector with mean 0 and covariance matrix \sigma^{2}I_{N} , where I_{N} is the N by N identity matrix. Let \gamma_{N} be the distribution or the law of the random vector {\mathbf{a}} . Then \gamma_{N} is just the centered Gaussian measure on \mathbb{R}^{N} , given by \gamma_{N}(G)=\frac{1}{(2\pi\sigma^{2})^{N/2}}\int_{G}e^{-\frac{|{\mathbf{a}}|^{2}}{2\sigma^{2}}}\,d{\mathbf{a}}\ {\rm for\ each\ Borel\ subset}\ G\subset\mathbb{R}^{N}, where |{\mathbf{a}}|=(\sum_{k:\lambda_{k}\leq n}a^{2}_{k})^{1/2} . For given 1\leq p,q\leq\infty , the average Nikolskii factor for \mathcal{P}_{n} is defined by N^{{\rm ave}}_{p,q}(\mathcal{P}_{n}):={E}\frac{\|P_{{\mathbf{a}}}\|_{q}}{\|P_{{\mathbf{a}}}\|_{p}}=\int_{\mathbb{R}^{N}}\frac{\|P_{{\mathbf{a}}}\|_{q}}{\|P_{{\mathbf{a}}}\|_{p}}d\gamma_{N}({\mathbf{a}})=\frac{1}{(2\pi\sigma^{2})^{N/2}}\int_{{R}^{N}}\frac{\|P_{{\mathbf{a}}}\|_{q}}{\|P_{{\mathbf{a}}}\|_{p}}e^{-\frac{|{\mathbf{a}}|^{2}}{2\sigma^{2}}}d{\mathbf{a}}. It follows immediately from the definitions of the two Nikolskii factors that for 1\leq p,q\leq\infty , N^{{\rm ave}}_{p,q}(\mathcal{P}_{n})\leq N^{{\rm wor}}_{p,q}(\mathcal{P}_{n}). This paper is devoted to investigating expected values of the Nikolskii factors of random diffusion polynomials. Given a polynomial basis p_{1},\dots,p_{N} , a random polynomial p is a polynomial whose coefficients are random, i.e., p=\sum_{k=1}^{N}X_{k}p_{k}, where X_{k},\,1\leq k\leq N are independent and identically distributed (i.i.d.) random variables. In recent years the study of random polynomials has attracted much interest, and there are numerous papers devoted to this field, see [ 2 , 3 , 4 , 6 , 7 , 8 , 11 , 12 , 14 , 21 , 29 , 30 , 31 , 33 , 36 , 37 , 38 , 40 , 42 ] . There are also many papers devoted to studying the Nikolskii type inequalities, see for example, [ 1 , 9 , 10 , 17 , 41 ] . However, as far as we know, there are only two papers devoted to investigating the average Nikolskii factors for random polynomials (see [ 30 , 31 ] ). Indeed, the authors in [ 30 ] studied the the average Nikolskii factors on the d -dimensional torus and obtained their exact orders, and later the authors in [ 31 ] studied the the average Nikolskii factors on the (d-1) -dimensional sphere and obtained their exact orders. It turns out that in many cases, the average Nikolskii factors on the d -dimensional torus and the (d-1) -dimensional sphere are significantly smaller than the worst Nikolskii factors. The purpose of this paper is to extend these results to the closed Riemannian manifold. We investigate the average Nikolskii factors