Comparison of persistent homologies for vector functions: from continuous to discrete and back (Q524581)

Let \(X\) be a triangulable space. One-parameter persistent homology studies the homological evolution of the sublevel sets of a (continuous) function \(f: X\to \mathbb{R}\). Similarly, in the multi-parameter generalization considered in this paper, one studies the homological evolution of the sublevel sets of a vector-valued function \(f: X\to \mathbb{R}^k\). In particular, the \(q\)-th rank invariant function associated to \(f\) is the function which associates to a pair \((a,b) \in \mathbb{R}^k \times \mathbb{R}^k\), \(a\leq b\), the rank of the linear morphism \(H_q(f^{-1}(-\infty, a]) \to H_q(f^{-1}(-\infty, b])\). Note that homology is always computed with field coefficients. In order to compare two rank invariants arising from different functions on possibly different spaces (but with the same codomain \(\mathbb{R}^k\)), the authors use the matching distance as introduced in [\textit{F. Cagliari} et al., Proc. Am. Math. Soc. 138, No. 8, 3003--3017 (2010; Zbl 1214.68424)]. The work in the current paper is motivated by the fact that one in practice needs to work with a simplicial approximation of \(X\), say in the form of a simplicial complex \(K\) and an approximation \(\phi: K\to \mathbb{R}^k\) of \(f\). Moreover, as the authors argue in the introduction, it is legitimate to assume that one only knows the value of \(\phi\) on the vertices of \(K\). The main contributions are: 1) Let \(V(K)\) denote the vertices of \(K\). Given \(\phi: V(K) \to \mathbb{R}^k\) one may extend \(\phi\) to the whole of \(K\) by means of axis-wise linear interpolation. This approach is known to introduce artifacts known as topological aliasing. The authors introduce a novel way of extending \(\phi\) which is free of such artifacts. 2) Let \(f: X\to \mathbb{R}^k\) and \(g: Y\to \mathbb{R}^k\), \(X\) and \(Y\) homeomorphic, with discrete approximations \(\phi: K\to \mathbb{R}^k\) and \(\theta: L\to \mathbb{R}^k\), respectively. The matching distance between the rank invariants of \(f\) and \(q\) can be well approximated by the matching distance between the rank invariants of \(\phi\) and \(\theta\). For a precise statement see Theorem 4.5. Some numerical evidence is provided towards the end.