Abstract:The standard convex closed hull of a set is defined as the intersection of all images, under the action of a group of rigid motions, of a half-space containing the given set. In this paper we propose a generalisation of this classical notion, that we call a $(K,\mathbb{H})$-hull, and which is obtained from the above construction by replacing a half-space with some other convex closed subset $K$ of the Euclidean space, and a group of rigid motions by a subset $\mathbb{H}$ of the group of invertible affine transformations. The main focus is put on the analysis of $(K,\mathbb{H})$-convex hulls of random samples from $K$.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to define and analyze the generalized convex hull of a set under the action of a broader transformation group. Specifically, the author proposes a concept called \((K,H)\)-convex hull, where \(K\) is a closed convex set in Euclidean space and \(H\) is a subset of affine transformations. The traditional convex hull is defined as the intersection of all half - spaces containing a given set, while the \((K,H)\)-convex hull is constructed by replacing these half - spaces with other closed convex sets \(K\) and using the transformations in \(H\).
### Main problems
1. **Definition of the generalized convex hull**:
- The author defines the \((K,H)\)-convex hull \(\text{conv}_{K,H}(A)\), which is the intersection of all images of \(K\) under the transformations in \(H\) that contain the set \(A\). That is:
\[
\text{conv}_{K,H}(A):=\bigcap_{(x,g)\in H:A\subseteq g(K + x)}g(K + x)
\]
- This definition allows for the study of more complex geometric structures and transformations, thus extending the concept of the traditional convex hull.
2. **\((K,H)\)-convex hull of random samples**:
- The paper also explores the properties of the \((K,H)\)-convex hull when \(A\) is a sample randomly drawn from \(K\). In particular, the asymptotic behavior of \(\text{conv}_{K,H}(\Xi_n)\) as the sample size \(n\) approaches infinity, where \(\Xi_n\) is a sample consisting of \(n\) independent and identically distributed random variables.
3. **Limit theorems**:
- The author proves that under certain conditions, the sequence of \((K,H)\)-convex hulls \(n\tilde{\log}(X_{K,H}(\Xi_n)\cap(\mathbb{R}^d\times V))\cap C_H\) converges in distribution to a random closed convex set \(\tilde{Z}_K\cap C_H\). Here, \(C_H\) is a closed convex cone and \(\tilde{Z}_K\) is a random set defined by a Poisson process.
### Key results
- **Basic properties**:
- The \((K,H)\)-convex hull is a closed convex set and satisfies some basic properties, such as idempotence.
- When \(H\) contains translation and scaling transformations, the \((K,H)\)-convex hull is consistent with the traditional convex hull.
- **Asymptotic behavior of random samples**:
- As the sample size increases, the asymptotic behavior of the \((K,H)\)-convex hull can be described by a Poisson process, which provides a theoretical basis for understanding the geometric structure of large - scale data.
- **Applications of limit theorems**:
- By studying the limit behavior of the \((K,H)\)-convex hull, some specific geometric and probabilistic conclusions can be derived. For example, in some cases, the limit form of the \((K,H)\)-convex hull can be simplified to a known geometric object.
### Conclusion
By introducing the concept of \((K,H)\)-convex hull, this paper provides a new method to study the convexity of sets, especially under the action of more complex transformation groups. Through the study of the asymptotic behavior of the \((K,H)\)-convex hull of random samples, the paper provides theoretical support for understanding the geometric structure of large - scale data.