Abstract:We first give a decomposition for a $T$-invariant sublinear expectation $\mathbb{E}=\sup_{P\in\Theta}\mathrm{E}_P$, and show that each component $\mathbb{E}^{(d)}=\sup_{P\in\Theta^{(d)}}\mathrm{E}_P$ of the decomposition has a finite period $p_d\in\mathbb{N}$, i.e., \[\mathbb{E}^{(d)}\left[f-f\circ T^{p_d}\right]=0, \quad f\in\mathcal{H}.\] Then we prove that a continuous invariant sublinear expectation that is strongly ergodic has a finite period $p_{\mathbb{E}}$, and each component $\Theta^{(d)}$ of its periodic decomposition is the convex hull of a finite set of $T^{p_d}$-ergodic probabilities.
As an application of the characterization, we prove an ergodicity result which shows that the limit of the $p_{\mathbb{E}}$-step time means achieves the upper expectation.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is about the periodic decomposition of Invariant Sublinear Expectations (ISE) and its properties under the strong ergodicity condition. Specifically, the author hopes to answer the question of how to achieve the upper and lower bound expectations through time means by conducting a more in - depth structural analysis of ISE.
### Main problems of the paper
1. **Periodic decomposition**:
- The author first gives a decomposition of a T - invariant sublinear expectation \( E=\sup_{P\in\Theta}E_P \) and proves that each decomposition component \( E^{(d)}=\sup_{P\in\Theta^{(d)}}E_P \) has a finite period \( p_d\in\mathbb{N} \), that is, it satisfies:
\[
E^{(d)}[f - f\circ T^{p_d}]=0,\quad\forall f\in H
\]
- Here, \( p_d \) is the smallest positive integer that makes the above equation hold, and is called the period of ISE \( E^{(d)} \).
2. **Strong ergodicity**:
- Next, the author proves that if a continuous invariant sublinear expectation has strong ergodicity, then it has a finite period \( p_E \), and each periodic decomposition component \( \Theta^{(d)} \) is the convex hull of a finite number of \( T^{p_d} \)-ergodic probabilities.
- As an application of this property, the author proves an ergodicity result, indicating that the limit of the \( p_E \)-step time means reaches the upper bound expectation.
3. **Relationship between time means and expectations**:
- The author further explores how to achieve the upper and lower bound expectations through time means. Specifically, for any probability measure \( P\in\Theta \) and bounded random variable \( \xi \), the author shows the following inequality:
\[
-E^{(1)}[-\xi]\leq\lim_{n\rightarrow\infty}\frac{1}{n}\sum_{k = 0}^{n - 1}\xi(T^k\omega)\leq E^{(1)}[\xi],\quad P - \text{a.s.}
\]
### Main conclusions
- **Existence of periodic decomposition**: The author proves that each T - invariant sublinear expectation can be decomposed into several periodic components.
- **Periodicity under strong ergodicity condition**: If a continuous T - invariant sublinear expectation is strongly ergodic, then it has a finite period, and each periodic decomposition component is the convex hull of a finite number of ergodic probabilities.
- **Limit behavior of time means**: For strongly ergodic ISE, the limit of its time means can reach the upper bound expectation, which provides a new perspective for understanding the law of large numbers under sublinear expectations.
### Application examples
The paper also provides some specific examples to illustrate the application of these theoretical results, including examples of independent and identically distributed random variable sequences and the existence of ISE periodic decomposition on countable state spaces.
In summary, by introducing the concept of periodic decomposition, this paper conducts an in - depth study of the structure and properties of invariant sublinear expectations, especially their performance under the strong ergodicity condition, thus providing new tools and methods for the development of sublinear expectation theory.