Random walks generated by affine mappings

Let G be a set of affine maps from \(R^ m\) into \(R^ m\) and for \(g\in G\) write \(g:x\to ax+b\). Define a Markov chain \(\{X_{\tau^ k}\}\) on \(R^ m\) by the recursion \(X_{n+1}=g_{n+1}X_ n\), \(n\geq 0\), where \(g_{n+1}\in G\) is distributed according to \(\mu_{X_ n}\), independent of \(X_ 0,X_ 1,...,X_{n-1}\). Introducing Khas'minskij's construction of an invariant measure based on the sampled chain \(\{X_{\tau \quad k}\}\) at its return times \(\tau^ k\), the author shows that \(\{X_ n\}\) is ergodic and has a unique stationary asymptotic distribution under the assumptions (in which some misprints are corrected): \((A_ 1)\) For any \(h\in C_ b(G)\) and \(x,y\in R^ m\), \(| E_ xh- E_ yh| \leq \phi (| x-y|)\| h\|\), where \(\phi:R_+\to R_+\) is a continuous function and \(\lim_{t\downarrow 0} \phi (t)\log^ 2(1/t)=0.\) \((A_ 2)\) There exist \(0<\epsilon_ 0<1\), \(\alpha >0\), such that \[ \sup_{x,| y|}E_ x[\log | ay| | \quad | ay| \geq \epsilon_ 0]\leq -\alpha. \] \((A_ 3)\) For every \(\epsilon >0\), there exist \(0<s<1\), such that \[ \inf_{| x-x'| \geq \epsilon}\Pr (| gx-g'x'| \leq s| x-x'|):=\beta >0. \] \((A_ 4)\) There exist \(0<\delta \leq 1\), \(k>0\), such that \[ \sup_{x,| y|}E_ x(\log^+| ay|)^{2+\delta}\leq k,\quad \sup_{x,| y|}E_ x(\log^+| b|)^{1+\delta}\leq k. \]

Recommendations

Cites work

Cited in

(11)

This page was built for publication: Random walks generated by affine mappings

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1102629)