Topological entropy on closed sets in \([0,1]^2\) (Q725662)

This paper generalizes the notion of topological entropy to closed subsets of $[0,1]^2$. A closed subset $R$ of $[0,1]^2$ is considered to be a set-valued function, where a point $x$ in the domain of $R$ maps to the (closed) set $\{y \in [0,1] \,:\, (x,y) \in R \}$. Such functions are upper-semicontinuous when viewed as maps from (some subset of) $[0,1]$ to the hyperspace $2^{[0,1]}$ of all closed subsets of $[0,1]$. Conversely, the graph of any such upper-semicontinuous function is a closed subset of $[0,1]^2$. \par The notion of topological entropy was first defined for self-maps of compact metric spaces. This paper introduces a version of that definition that still makes sense in the context of set-valued self-maps of the unit interval. \par It turns out that this generalized notion of topological entropy exhibits several interesting features. For example, the topological entropy of a set-valued function can depend entirely on a finite closed subset of its graph. This is impossible with normal (single-valued) functions $[0,1] \rightarrow [0,1]$, and occurs in set-valued functions due to a sort of ``feedback look'' that arises when single inputs produce multiple outputs. \par For another example, the authors show that it is possible for a set-valued map $R \subseteq [0,1]^2$ to have zero entropy, while adding a single extra point to $R$ produces set-valued map with infinite entropy. \par Importantly, the authors show that the calculation of the topological entropy of a set-valued map can in some cases be reduced to a combinatorial procedure that involves counting intersecting grid boxes. This is important because the definition of topological entropy is complicated, and it is not clear from the definition how one might go about actually computing it. \par In short, the paper contains many results on the topological entropy of set-valued maps. It is worth noting that some of these results were anticipated by Kelly and Tennant in their work on the entropy of set-valued maps. The authors are careful to note this overlap when it arises. The paper is of interest to anyone working on set-valued functions and their associated inverse limits.