Complexity for extended dynamical systems (Q930558)

A special class of extended dynamical systems is considered. Such a dynamical system consists of a set \(\mathcal F\) of real functions with real arguments endowed with a translation invariant metric and a probability measure which is invariant and ergodic with respect to the action exerted on \(\mathcal F\) by the space translation and by a flow \(\varphi\) of time evolution. Several assumptions are made on the set \(\mathcal F\) and the flow \(\varphi\). Under these assumptions a notion of topological entropy for \(\varphi\) was introduced by \textit{P. Collet} and \textit{J.-P. Eckmann} [Nonlinearity 12, No. 3, 451--473 (1999; Zbl 0984.37095), Nonlinearity 14, No. 4, 907 (2001; Zbl 1143.37328)]. It is the aim of the authors to find a measure of the complexity of the \(\varphi\)-action on \(\mathcal F\). Certain hypotheses which are required by this complexity are proposed and proved to be verified by the Kolmogorov complexity [cf. \textit{M. Li} and \textit{P. M. B. Vitányi}, Texts in Computer Science, New York, NY: Springer (2008; Zbl 1185.68369)]. Two main results are presented. The first one shows that one can define a complexity per unit time and unit volume. The second one is a variational principle for the complexity which proves that the supremum of the complexity for functions in the supports of the invariant measures of the system coincides with the supremum of the metric entropies along space time boxes of bounded aspect ratio.