The theory of multidimensional persistence (Q1028586): Difference between revisions

By extending the usual definition of a filtered space as a collection of subspaces \(\{X_v\}_v\) indexed by natural numbers (and satisfying containment conditions) to allow for indexing by \(n\)-tuples of natural numbers, the authors extend the definition of persistent homology of a filtered simplicial complex to that of a \textit{multifiltered} one. An attempt is then made to extend the constructions used in their previous work [Discrete Comput. Geom. 33, No.~2, 249--274 (2005; Zbl 1069.55003)]: the induced multifiltration of homology modules \(\{H_\ast(X_v)\}_v\) is identified with a single module over a polynomial ring in \(n\) variables, and the resulting module is then classified and parametrized. It is here the authors run aground: they identify such modules with orbits in a quasi-projective variety acted on by an algebraic group; however, this does not readily yield a complete discrete invariant; i.e., an invariant that is independent of the coefficients chosen and that captures all of the homological information of the multifiltration. Changing course, the authors move on to define a discrete invariant (the \textit{rank invariant}), given in terms of the ranks of the inclusion maps \(H_\ast(X_v)\rightarrow H_\ast(X_w)\) for \(X_v\subseteq X_w\); and although this reproduces the usual barcode of persistent homology for a filtration, it does not extend to a complete invariant for a multifiltration. The authors assume the reader has a passing acquaintance with homological algebra and algebraic geometry; familiarity with persistent homology is helpful. The reader should be aware, however, that the article suffers from a lack of editorial input: notation is inconsistent and occasionally conflicting, and the section on parametrization lacks a clarity of exposition.