An Iterated Random Function with Lipschitz Number One (Q4442903)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 2030774
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | An Iterated Random Function with Lipschitz Number One |
scientific article; zbMATH DE number 2030774 |
Statements
An Iterated Random Function with Lipschitz Number One (English)
0 references
21 January 2004
0 references
iterated random function
0 references
Markov process
0 references
stationary distribution
0 references
0.8064392
0 references
0.80631435
0 references
0.7939653
0 references
0.7938986
0 references
0.79136264
0 references
In a recent paper, \textit{P. Diaconis} and \textit{D. Freedman} [SIAM Rev. 41, 45--76 (1999; Zbl 0926.60056)] described a general method for studying Markov chains. This method consists of viewing a state transition of the chain as the action of a function chosen at random from a family of state transformations. The Markov process can then be expressed as an iteration of these functions. When the state space is a metric space, e.g. \(\mathbb{R}\), they show that if the functions exhibit a certain average contractivity, then running the iteration backward rather than forward produces geometric convergence to a unique stationary distribution for the chain. An example of a Markov process originally proposed by \textit{G. Letac} [in: Random matrices and their applications. Contemp. Math. 50, 263--273 (1986; Zbl 0587.60057)] is defined on the nonnegative reals \(\mathbb{R}^+= [0,+\infty)\) by starting with a point \(x_0\in \mathbb{R}^+\) and for \(k= 0,1,\dots\) setting \[ x_{k+1}= f_{\theta_{k+ 1}} (x_k)\equiv |\theta_{k+1}- x_k|,\tag{1} \] where \(\theta_k \in \mathbb{R}^+\) are chosen independently from a fixed distribution \(\mu\) (on \(\mathbb{R}^+\)). The authors prove Letac's conjecture for compactly supported \(\mu\). Theorem 1. Let \(\mu\) be a distribution on a bounded interval \(Y\subset \mathbb{R}^+\) with \(\mu\) not supported on a lattice. Then \(\pi_\mu\) is a unique stationary distribution for the process (1). Moreover, if \(x_0\) is chosen according to an initial distribution \(\pi_0\), then the distribution \(\pi_0\) of \(x_0\) converges to \(\pi_p\) on the weak topology. The authors prove this theorem in Sections 2 and 3, and in Sections 4 and 5 they analyze the special case of \(\mu\) supported on just two points. Also the authors learned of a paper by \textit{J. C. Mattingly} [Theory Probab. Appl. 47, 333--343 (2002) and Teor. Veroyatn. Primen 47, 388--397 (2002; Zbl 1032.60064)] addressing related questions, that the paper contains results similar to ours.
0 references