Abstract:In this paper, we generalize the result on the average volume of random polytopes with vertices following beta distributionsto the case of non-identically distributed vectors. Specifically,we consider the convex hull of independent random vectors in $\mathbb{R}^d$, where each vector follows a beta distribution with potentially different parameters. We derive an expression for the expected volume of these generalized beta--polytopes. Additionally, we compute the expected value of a functional introduced by Wieacker, which involves the distance of facets from the origin and their volumes.Our results extend the findings of Kabluchko, Temesvari,and Thäle. Key techniques used in the proofs include the Blaschke--Petkantschin formula, Kubota's formula, and projections of beta distributed random vectors.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is about the calculation of the average volume of mixed random beta - polytopes. Specifically, the author hopes to generalize the existing results to the case of non - identically distributed vectors.
### Background and Problem Description
In previous literature, Kabluchko et al. have studied the average volume of random polytopes whose vertices follow the same Beta distribution. However, in practical applications, different vertices may follow different Beta distributions, so a more general formula is required to handle this situation.
### Main Contributions
1. **Generalization to the Non - Identically Distributed Case**:
- The author considers independent random vectors \(X_1,\ldots,X_n\) in \(\mathbb{R}^d\) space, and each vector follows a different Beta distribution.
- Derives the expression for the expected volume of these generalized Beta polytopes.
2. **Expected Value of the Wieacker Function**:
- Calculates the expected value of the function introduced by Wieacker, which involves the distance from the face to the origin and its volume.
### Formula Derivation
For non - identically distributed Beta polytopes, the author's main results are as follows:
#### Theorem 2.1: Expected Volume
Given the parameter vector \(\boldsymbol{\beta}=(\beta_1,\ldots,\beta_n)\) with \(\beta_i > - 1\), let \(X_1,\ldots,X_n\) be independent points in \(\mathbb{R}^d\), and each point follows the density function \(f_{d,\beta_i}\) respectively. Define the random polytope \(P_{n,d}^{\boldsymbol{\beta}}=\text{conv}(X_1,\ldots,X_n)\), then its expected volume is:
\[
E[V_d(P_{n,d}^{\boldsymbol{\beta}})]=\frac{\kappa_d}{\pi^{d + 1}2^{2d}}\sum_{i_1 < \cdots < i_{d+1}}\left(\frac{(d + 1)^2}{2}+\sum_{k = 1}^{d+1}\beta_{i_k}\right)^{d+1}\prod_{k = 1}^{d+1}\frac{\Gamma\left(\frac{d+2}{2}+\beta_{i_k}\right)}{\Gamma\left(\frac{d+3}{2}+\beta_{i_k}\right)}
\]
\[
\times\int_{-1}^{1}(1 - h^2)^{\sum_{k = 1}^{d+1}\beta_{i_k}+\frac{d^2+2d - 1}{2}}\prod_{k\neq i_1,\ldots,i_{d+1}}F_{\beta_k+\frac{d - 1}{2}}(h)\,dh
\]
#### Theorem 2.2: Expected Value of the Wieacker Function
For the function \(T_{a,b}^{d-1,d}(P)=\sum_{F\in\mathcal{F}_{d-1}(P)}\text{dist}^a(F,0)\text{Vol}_b(F)\) introduced by Wieacker, its expected value is:
\[
E[T_{a,b}^{d-1,d}(P_{n,d}^{\boldsymbol{\beta}})]=d!\kappa_d\sum_{i_1 < \cdots < i_d}c_{d,\beta_i}c_{d-1,\beta_i}E[\text{Vol}_{d-1}(\text{conv}(X_{i_1},\ldots,X_{i_d}))]^{b+1}
\]