Entropy formulas for dynamical systems with mistakes (Q1760092)

The authors study measure-theoretic recurrence properties of topological dynamical systems given by continuous selfmaps \(f:X\rightarrow X\) of compact metric spaces. They extend classical results about the exponential growth rates of first and minimal return times to dynamical balls and their relations to metric entropy. More precisely, they replace the dynamical balls (balls in the Bowen-Dinaburg metrics) by so-called mistake dynamical balls \[ B_n(g;x,\varepsilon) = \bigcup_{\Lambda \in I(g;n,\varepsilon)}\left\{ y\in X\;:\;d(f^i(x),f^i(y)) < \varepsilon \mathrm{ for all } i\in\Lambda\right\}, \] where \(I(g;n,\varepsilon) = \{\Lambda \subset [0,n-1]\cap\mathbb{N} : \#\Lambda \geq n - g(n,\varepsilon)\}\) and \(g:\mathbb{N}\times(0,\varepsilon_0]\rightarrow\mathbb{N}\) is a function which increases with respect to \(n\) with sublinear growth, called a mistake function. The motivation for looking at such mistake balls consists in errors which may arise by the modelled dynamical process itself due to disturbances, or by the evaluation of the orbits. For every mistake function \(g\), the first return time to \(B_n(g;x,\varepsilon)\), denoted by \(R_n(g;x,\varepsilon)\), is the least positive integer \(k\) with \(f^k(x) \in B_n(g;x,\varepsilon)\), and the first main result of the paper (Theorem A) asserts that for every ergodic measure \(\mu\) and every mistake function \(g\) the entropy is given by \[ h_{\mu}(f) = \lim_{\varepsilon\searrow0} \limsup_{n\rightarrow\infty}\frac{1}{n}\log R_n(g;x,\varepsilon) \] for \(\mu\)-almost every \(x\in X\), where the \(\limsup\) can be replaced by \(\liminf\). A generalization for non-ergodic invariant measures is provided. Using weighted return times, one obtains an analogous result for the metric pressure with respect to a continuous observable. In addition to the first return time, also the minimal return time \(S_n(g;x,\varepsilon)\) to a mistake ball, defined as the least positive integer \(k\) with \(f^{-k}(B_n(g;x,\varepsilon)) \cap B_n(g;x,\varepsilon) \neq \emptyset\), is considered. Using a modified version of Bowen's specification property in terms of mistake functions, the authors prove their second main result (Theorem B), which states that for a system with the modified specification property and positive entropy \(h_{\mu}(f)\) for an ergodic measure \(\mu\), for \(\mu\)-almost every \(x\in X\) \[ \lim_{\varepsilon\searrow0}\limsup_{n\rightarrow\infty} \frac{1}{n}S_n(g;x,\varepsilon) = 1. \] The third result (Theorem C) relates the exponential growth rates of weighted first and minimal return times for subshifts of finite type to the entropy with respect to the unique equilibrium state and the topological pressure with respect to a Hölder continuous potential. Finally, the authors prove a formula for the metric entropy of a suspension semiflow whose base flow satisfies the mistake specification property, in terms of first and minimal return times (Theorem D). As examples, the authors apply their results to \(\beta\)-transformations and Axiom A flows.