Orthogonal M\"obius Inversion and Grassmannian Persistence Diagrams

Aziz Burak Gülen,Facundo Mémoli,Zhengchao Wan

arXiv (Cornell University)(2023)

引用 0|浏览0
暂无评分
摘要
We introduce the notion of Orthogonal M\"obius Inversion, a custom-made analog to M\"obius inversion on the poset of intervals $\mathsf{Int}(P)$ of a linear poset $P$. This notion is applicable to functions $\bar{\mathsf{F}}:\mathsf{Int}(P)\to\mathsf{Gr}(V)$ whose target space is the Grassmannian of a given inner product space $V$. When this notion of inversion is applied to such a function $\bar{\mathsf{F}}$, we obtain another function $\mathsf{OI}\bar{\mathsf{F}}:\mathsf{Int}(P)\to\mathsf{Gr}(V)$, the Grassmannian persistence diagram (GPD) induced by $\bar{\mathsf{F}}$, which assigns to every interval a subspace of $V$ by "compressing" the information contained in $\bar{\mathsf{F}}$. Given a $1$-parameter filtration $\mathsf{F}$ of a finite simplicial complex $K$, and an inner product on the $q$-th chain group $C_q^K$ of $K$, we consider two chain group invariants (as functions $\bar{\mathsf{F}}:\mathsf{Int}(P) \to \mathsf{Gr}(C_q^K)$): birth-death spaces, and the space of harmonic persistent cycles (i.e. persistent Laplacian kernels). By applying Orthogonal M\"obius Inversion to these invariants (with respective compatible choices of partial orders on $\mathsf{Int}(P)$), we obtain two, a priori, different GPDs associated to the filtration $\mathsf{F}$. Each of these two particular GPDs assigns a vector subspace of the $q$-th chain group to each interval in the classical persistence diagram. Interestingly, we prove that not only do these two GPDs coincide on non-ephemeral intervals but that they also both refine the standard persistence diagram in the sense that (1) all supports coincide (away from the diagonal) and (2) GPDs carry extra information in terms of cycle representatives. We also show that Orthogonal M\"obius Inversion satisfies a certain type of functoriality. As a result of this functoriality, we prove that GPDs are stable with respect to a suitable notion of distance.
更多
查看译文
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要