Total variation bound for Hadwiger's functional using Stein's method
For a convex body \(K\subset{\mathbb R}^d\), the Steiner formula leads to the intrinsic volumes \(\nu_k(K)\) and these appear in the Wills functional \(W(K)= \sum_{k=0}^d \nu_k(K)\). According to \textit{H. Hadwiger} [Monatsh. Math. 79, 213--221 (1975; Zbl 0305.52007)], the latter can be represented by \(W(K)=\int_{{\mathbb R}^d} e^{-\pi\,\operatorname{dist}^2(x,K)}\,dx\), where \(\operatorname{dist}(\cdot,K)\) is the distance to \(K\). This gives rise to a discrete random variable \(V_K\) (on some probability space \(\Omega\)) with \({\mathbb P}(V_K=k)=\nu_k(K)/W(K)\), a random vector \(X_K\) with density proportional to \(e^{-\pi\,\operatorname{dist}^2(\cdot,K)}\), and a real random variable \(H_K=\pi\,\operatorname{dist}^2(X_K,K)\). The authors explain how the probabilistic properties of \(H_K\) can be transferred to \(V_K\) and vice versa, therefore they deal with \(H_K\), but they also explain why the obtained central limit theorem for \(H_K\) is difficult to transfer to \(V_K\). For their main result, they define the random variable \N\[\NF_K= \frac{\operatorname{dist}^2(X_K,K)-{\mathbb E}[\operatorname{dist}^2(X_K,K)]}{{\sqrt{\mathbb{V}\mathrm{ar}(\operatorname{dist}^2(X_K,K))}}}\N\]\Nand prove for all \(\alpha\ge 0\) and all \(c>0\) that \N\[\N\sup_{d\ge 1} \sup_{K:\,\dim(K)\le c d^\alpha} d^{\frac{1}{2}-\alpha} d_{TV}(F_K,N)<\infty.\N\]\NHere \(d_{TV}\) is the total variation distance and \(N\) is a standard Gaussian random variable. A consequence is the following central limit theorem. If \(\alpha\in [0,\frac{1}{2})\), if \((d_n)\) is a sequence of integers diverging to infinity and if \((K_n)\) is a sequence of convex bodies satisfying \(K_n\subset{\mathbb R}^{d_n}\) and \(\operatorname{diam}(K_n)=O(d_n^\alpha)\), then \(F_{K_n}\stackrel{\operatorname{law}}{\longrightarrow} {\mathcal N}(0,1)\) as \(n\to\infty\), where \({\mathcal N}(0,1)\) denotes the standard normal distribution. The proof employs Stein's method; another tool is the Brascamp-Lieb inequality. Also, the ultra log-concavity of the sequence of intrinsic volumes plays a role.
- A discrete log-Sobolev inequality under a Bakry-Émery type condition
- Approximation of convex bodies by polytopes
- Asymptotic development for the CLT in total variation distance
- Concentration of the intrinsic volumes of a convex body
- Das Wills'sche Funktional
- Fluctuations of eigenvalues and second order Poincaré inequalities
- Gaussian phase transitions and conic intrinsic volumes: steining the Steiner formula
- scientific article; zbMATH DE number 5152130 (Why is no real title available?)
- Normal approximations with Malliavin calculus. From Stein's method to universality
- On extensions of the Brunn-Minkowski and Prekopa-Leindler theorems, including inequalities for log concave functions, and with an application to the diffusion equation
- Processus gaussiens et volumes mixtes
- Stein's method for comparison of univariate distributions
- Zur Gitterpunktanzahl konvexer Mengen
This page was built for publication: Total variation bound for Hadwiger's functional using Stein's method
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6860808)
