Bounding the interleaving distance for mapper graphs with a loss function (Q6933054)

In this paper, the authors introduce a new method for comparing similarity betwen between data sets consisting of a graph \(G\) and a map \(f: G \rightarrow \mathbb{R}^d\). In particular, using the mapper graph as a discretisation of these data sets. Here, natural transformations can be used to compare the functorial representations, and then obtain a measure of (dis)similarity.\N\NUnfortunately, the diagrams involved do not always commute. To overcome this, the author's method uses families of maps that are not quite natural transformations to measure how far away the diagrams in an interleaving are from commuting.\N\NIn particular, they show that the computations involved have polynomial time given certain input information.