Searcharxiv⌕ Search

arXiv subjects

Facundo Mémoli

Publications and source records attributed to Facundo Mémoli.

At least 55 records · Page 3Linked to original sources
arXiv2022

Extracting Persistent Clusters in Dynamic Data via Möbius inversion

Identifying and representing clusters in time-varying network data is of particular importance when studying collective behaviors emerging in nature, in mobile device networks or in social networks. Based on combinatorial, categorical, and persistence theoretic viewpoints, we establish a stable functorial pipeline for the summarization of the evolution of clusters in a time-varying network. We first construct a complete summary of the evolution of clusters in a given time-varying network over a set of entities $X$ of which takes the form of a formigram. This formigram can be understood as a certain Reeb graph $\mathcal{R}$ which is labeled by subsets of $X$. By applying Möbius inversion to the formigram in two different manners, we obtain two dual notions of diagram: the maximal group diagram and the persistence clustergram, both of which are in the form of an `annotated' barcode. The maximal group diagram consists of time intervals annotated by their corresponding maximal groups -- a notion due to Buchin et al., implying that we recognize the notion of maximal groups as a special instance of generalized persistence diagram by Patel. On the other hand, the persistence clustergram is mostly obtained by annotating the intervals in the zigzag barcode of the Reeb graph $\mathcal{R}$ with certain merging/disbanding events in the given time-varying network. We show that both diagrams are complete invariants of formigrams (or equivalently of trajectory grouping structure by Buchin et al.) and thus contain more information than the Reeb graph $\mathcal{R}$.

math.AT↗
arXiv2022

Weisfeiler-Lehman meets Gromov-Wasserstein

The Weisfeiler-Lehman (WL) test is a classical procedure for graph isomorphism testing. The WL test has also been widely used both for designing graph kernels and for analyzing graph neural networks. In this paper, we propose the Weisfeiler-Lehman (WL) distance, a notion of distance between labeled measure Markov chains (LMMCs), of which labeled graphs are special cases. The WL distance is polynomial time computable and is also compatible with the WL test in the sense that the former is positive if and only if the WL test can distinguish the two involved graphs. The WL distance captures and compares subtle structures of the underlying LMMCs and, as a consequence of this, it is more discriminating than the distance between graphs used for defining the state-of-the-art Wasserstein Weisfeiler-Lehman graph kernel. Inspired by the structure of the WL distance we identify a neural network architecture on LMMCs which turns out to be universal w.r.t. continuous functions defined on the space of all LMMCs (which includes all graphs) endowed with the WL distance. Finally, the WL distance turns out to be stable w.r.t. a natural variant of the Gromov-Wasserstein (GW) distance for comparing metric Markov chains that we identify. Hence, the WL distance can also be construed as a polynomial time lower bound for the GW distance which is in general NP-hard to compute.

cs.LG↗
arXiv2021

On $p$-metric spaces and the $p$-Gromov-Hausdorff distance

For each given $p\in[1,\infty]$ we investigate certain sub-family $\mathcal{M}_p$ of the collection of all compact metric spaces $\mathcal{M}$ which are characterized by the satisfaction of a strengthened form of the triangle inequality which encompasses, for example, the strong triangle inequality satisfied by ultrametric spaces. We identify a one parameter family of Gromov-Hausdorff like distances $\{d_{\mathrm{GH}}^{\scriptscriptstyle{(p)}}\}_{p\in[1,\infty]}$ on $\mathcal{M}_p$ and study geometric and topological properties of these distances as well as the stability of certain canonical projections $\mathfrak{S}_p:\mathcal{M}\rightarrow \mathcal{M}_p$. For the collection $\mathcal{U}$ of all compact ultrametric spaces, which corresponds to the case $p=\infty$ of the family $\mathcal{M}_p$, we explore a one parameter family of interleaving-type distances and reveal their relationship with $\{d_{\mathrm{GH}}^{\scriptscriptstyle{(p)}}\}_{p\in[1,\infty]}$.

math.MG↗
arXiv2021

The Gromov-Hausdorff distance between ultrametric spaces: its structure and computation

The Gromov-Hausdorff distance ($d_\mathrm{GH}$) provides a natural way of quantifying the dissimilarity between two given metric spaces. It is known that computing $d_\mathrm{GH}$ between two finite metric spaces is NP-hard, even in the case of finite ultrametric spaces which are highly structured metric spaces in the sense that they satisfy the so-called \emph{strong triangle inequality}. Ultrametric spaces naturally arise in many applications such as hierarchical clustering, phylogenetics, genomics, and even linguistics. By exploiting the special structures of ultrametric spaces, (1) we identify a one parameter family $\{d_\mathrm{GH}^{(p)}\}_{p\in[1,\infty]}$ of distances defined in a flavor similar to the Gromov-Hausdorff distance on the collection of finite ultrametric spaces, and in particular $d_\mathrm{GH}^{(1)} =d_\mathrm{GH}$. The extreme case when $p=\infty$, which we also denote by $u_\mathrm{GH}$, turns out to be an ultrametric on the collection of ultrametric spaces. Whereas for all $p\in[1,\infty)$, $d_\mathrm{GH}^{(p)}$ yields NP-hard problems, we prove that surprisingly $u_\mathrm{GH}$ can be computed in polynomial time. The proof is based on a structural theorem for $u_\mathrm{GH}$ established in this paper; (2) inspired by the structural theorem for $u_\mathrm{GH}$, and by carefully leveraging properties of ultrametric spaces, we also establish a structural theorem for $d_\mathrm{GH}$ when restricted to ultrametric spaces. This structural theorem allows us to identify special families of ultrametric spaces on which $d_\mathrm{GH}$ is computationally tractable. These families are determined by properties related to the doubling constant of metric space. Based on these families, we devise a fixed-parameter tractable (FPT) algorithm for computing the exact value of $d_\mathrm{GH}$ between ultrametric spaces. We believe ours is the first such algorithm to be identified.

math.MG↗
arXiv2021

Robust Hierarchical Clustering for Directed Networks: An Axiomatic Approach

We provide a complete taxonomic characterization of robust hierarchical clustering methods for directed networks following an axiomatic approach. We begin by introducing three practical properties associated with the notion of robustness in hierarchical clustering: linear scale preservation, stability, and excisiveness. Linear scale preservation enforces imperviousness to change in units of measure whereas stability ensures that a bounded perturbation in the input network entails a bounded perturbation in the clustering output. Excisiveness refers to the local consistency of the clustering outcome. Algorithmically, excisiveness implies that we can reduce computational complexity by only clustering a subset of our data while theoretically guaranteeing that the same hierarchical outcome would be observed when clustering the whole dataset. In parallel to these three properties, we introduce the concept of representability, a generative model for describing clustering methods through the specification of their action on a collection of networks. Our main result is to leverage this generative model to give a precise characterization of all robust -- i.e., excisive, linear scale preserving, and stable -- hierarchical clustering methods for directed networks. We also address the implementation of our methods and describe an application to real data.

cs.LG↗
arXiv2021

The ultrametric Gromov-Wasserstein distance

In this paper, we investigate compact ultrametric measure spaces which form a subset $\mathcal{U}^w$ of the collection of all metric measure spaces $\mathcal{M}^w$. Similar as for the ultrametric Gromov-Hausdorff distance on the collection of ultrametric spaces $\mathcal{U}$, we define ultrametric versions of two metrics on $\mathcal{U}^w$, namely of Sturm's distance of order $p$ and of the Gromov-Wasserstein distance of order $p$. We study the basic topological and geometric properties of these distances as well as their relation and derive for $p=\infty$ a polynomial time algorithm for their calculation. Further, several lower bounds for both distances are derived and some of our results are generalized to the case of finite ultra-dissimilarity spaces.

math.MG↗
arXiv2021

Characterization of Gromov-type geodesics

The collection $\mathcal{M}$ of all isometry classes of compact metric spaces endowed with the Gromov-Hausdorff distance $d_\mathcal{GH}$ is known to be a geodesic space. However, there is no known structural characterization of geodesics in $\mathcal{M}$. In this paper we provide two such characterizations. We first prove that every Gromov-Hausdorff geodesic is in fact a geodesic in the Hausdorff hyperspace of some compact metric space, which we call a Hausdorff geodesic. Inspired by this characterization, we further elucidate a structural connection between Hausdorff geodesics and Wasserstein geodesics: every Hausdorff geodesic is equivalent to a so-called Hausdorff displacement interpolation. This equivalence allows us to establish that every Gromov-Hausdorff geodesic is dynamic, a notion which we develop in analogy with dynamic optimal couplings in the theory of optimal transport. Besides geodesics in $\mathcal{M}$, we also study geodesics on the collection $\mathcal{M}^w$ of isomorphism classes of compact metric measure spaces. Sturm constructed a family of Gromov-type distances on $\mathcal{M}^w$, which we denote $d_{\mathcal{GW},{p}}^\mathrm{S}$ (for $p\in[1,\infty)$), and proved that $(\mathcal{M}^w,d_{\mathcal{GW},{p}}^\mathrm{S})$ is also a geodesic space. We are interested in $d_{\mathcal{GW},{p}}^\mathrm{S}$ geodesics which are (essentially) Wasserstein geodesics. We prove the set of such geodesics is dense in the set of all $d_{\mathcal{GW},{p}}^\mathrm{S}$ geodesics and identify a rich class of such geodesics.

math.MG↗
arXiv2020

The Gaussian Transform

We introduce the Gaussian transform (GT), an optimal transport inspired iterative method for denoising and enhancing latent structures in datasets. Under the hood, GT generates a new distance function (GT distance) on a given dataset by computing the $\ell^2$-Wasserstein distance between certain Gaussian density estimates obtained by localizing the dataset to individual points. Our contribution is twofold: (1) theoretically, we establish firstly that GT is stable under perturbations and secondly that in the continuous case, each point possesses an asymptotically ellipsoidal neighborhood with respect to the GT distance; (2) computationally, we accelerate GT both by identifying a strategy for reducing the number of matrix square root computations inherent to the $\ell^2$-Wasserstein distance between Gaussian measures, and by avoiding redundant computations of GT distances between points via enhanced neighborhood mechanisms. We also observe that GT is both a generalization and a strengthening of the mean shift (MS) method, and it is also a computationally efficient specialization of the recently proposed Wasserstein Transform (WT) method. We perform extensive experimentation comparing their performance in different scenarios.

cs.LG↗
arXiv2019

The Gromov-Wasserstein distance between networks and stable network invariants

We define a metric---the network Gromov-Wasserstein distance---on weighted, directed networks that is sensitive to the presence of outliers. In addition to proving its theoretical properties, we supply network invariants based on optimal transport that approximate this distance by means of lower bounds. We test these methods on a range of simulated network datasets and on a dataset of real-world global bilateral migration. For our simulations, we define a network generative model based on the stochastic block model. This may be of independent interest for benchmarking purposes.

cs.DM↗
arXiv2019

The reflection distance between zigzag persistence modules

By invoking the reflection functors introduced by Bernstein, Gelfand, and Ponomarev in 1973, in this paper we define a metric on the space of all zigzag modules of a given length, which we call the reflection distance. We show that the reflection distance between two given zigzag modules of the same length is an upper bound for the $\ell^1$-bottleneck distance between their respective persistence diagrams.

math.AT↗
arXiv2018

A Topological Study of Functional Data and Fréchet Functions of Metric Measure Spaces

We study the persistent homology of both functional data on compact topological spaces and structural data presented as compact metric measure spaces. One of our goals is to define persistent homology so as to capture primarily properties of the shape of a signal, eliminating otherwise highly persistent homology classes that may exist simply because of the nature of the domain on which the signal is defined. We investigate the stability of these invariants using metrics that downplay regions where signals are weak. The distance between two signals is small if they exhibit high similarity in regions where they are strong, regardless of the nature of their full domains, in particular allowing different homotopy types. Consistency and estimation of persistent homology of metric measure spaces from data are studied within this framework. We also apply the methodology to the construction of multi-scale topological descriptors for data on compact Riemannian manifolds via metric relaxations derived from the heat kernel.

math.AT↗
arXiv2018

Sketching and Clustering Metric Measure Spaces

Two important optimization problems in the analysis of geometric data sets are clustering and sketching. Here, clustering refers to the problem of partitioning some input metric measure space (mm-space) into k clusters, minimizing some objective function f. Sketching, on the other hand, is the problem of approximating some mm-space by a smaller one supported on a set of k points. Specifically, we define the k-sketch of some mm-space M to be the nearest neighbor of M in the set of k-point mm-spaces, under some distance function ρon the set of mm-spaces. In this paper, we demonstrate a duality between general classes of clustering and sketching problems. We present a general method for efficiently transforming a solution for a clustering problem to a solution for a sketching problem, and vice versa, with approximately equal cost. More specifically, we obtain the following results. 1. For metric spaces, we consider the case where the clustering objective is minimizing the maximum cluster diameter. We show that the ratio between the sketching and clustering objectives is constant over compact metric spaces. 2. We extend these results to the setting of metric measure spaces where we prove that the ratio of sketching to clustering objectives is bounded both above and below by some universal constants. In this setting, the clustering objective involves minimizing various notions of the l_p-diameters} of the clusters. 3. We consider two competing notions of sketching for mm-spaces, with one of them being more demanding than the other. These notions arise from two different definitions of p-Gromov-Wasserstein distance that have appeared in the literature. We then prove that whereas the gap between these can be arbitrarily large, in the case of doubling metric spaces the resulting sketching objectives are polynomially related.

cs.CG↗
arXiv2018

Explicit Geodesics in Gromov-Hausdorff Space

We provide an alternative, constructive proof that the collection $\mathcal{M}$ of isometry classes of compact metric spaces endowed with the Gromov-Hausdorff distance is a geodesic space. The core of our proof is a construction of explicit geodesics on $\mathcal{M}$. We also provide several interesting examples of geodesics on $\mathcal{M}$, including a geodesic between $\mathbb{S}^0$ and $\mathbb{S}^n$ for any $n\geq 1$.

math.MG↗
arXiv2018

Quantitative Simplification of Filtered Simplicial Complexes

We introduce a new invariant defined on the vertices of a given filtered simplicial complex, called codensity, which controls the impact of removing vertices on persistent homology. We achieve this control through the use of an interleaving type of distance between fitered simplicial complexes. We study the special case of Vietoris-Rips filtrations and show that our bounds offer a significant improvement over the immediate bounds coming from considerations related to the Gromov-Hausdorff distance. Based on these ideas we give an iterative method for the practical simplification of filtered simplicial complexes. As a byproduct of our analysis we identify a notion of core of a filtered simplicial complex which admits the interpretation as a minimalistic simplicial filtration which retains all the persistent homology information.

math.AT↗
arXiv2018

The Distortion of the Reeb Quotient Map on Riemannian Manifolds

Given a metric space $X$ and a function $f: X \to \mathbb{R}$, the Reeb construction gives metric a space $X_f$ together with a quotient map $X \to X_f$. Under suitable conditions $X_f$ becomes a metric graph and can therefore be used as a graph approximation to $X$. The Gromov-Hausdorff distance from $X_f$ to $X$ is bounded by the half of the metric distortion of the quotient map. In this paper we consider the case where $X$ is a compact Riemannian manifold and $f$ is an excellent Morse function. In this case we provide bounds on the distortion of the quotient map which involve the first Betti number of the original space and a novel invariant which we call thickness.

math.MG↗
arXiv2018

A functorial Dowker theorem and persistent homology of asymmetric networks

We study two methods for computing network features with topological underpinnings: the Rips and Dowker persistent homology diagrams. Our formulations work for general networks, which may be asymmetric and may have any real number as an edge weight. We study the sensitivity of Dowker persistence diagrams to asymmetry via numerous theoretical examples, including a family of highly asymmetric cycle networks that have interesting connections to the existing literature. In particular, we characterize the Dowker persistence diagrams arising from asymmetric cycle networks. We investigate the stability properties of both the Dowker and Rips persistence diagrams, and use these observations to run a classification task on a dataset comprising simulated hippocampal networks. Our theoretical and experimental results suggest that Dowker persistence diagrams are particularly suitable for studying asymmetric networks. As a stepping stone for our constructions, we prove a functorial generalization of a theorem of Dowker, after whom our constructions are named.

math.AT↗
arXiv2018

Reeb Posets and Tree Approximations

A well known result in the analysis of finite metric spaces due to Gromov says that given any $(X,d_X)$ there exists a \emph{tree metric} $t_X$ on $X$ such that $\|d_X-t_X\|_\infty$ is bounded above by twice $\mathrm{hyp}(X)\cdot \log(2\,|X|)$. Here $\mathrm{hyp}(X)$ is the \emph{hyperbolicity} of $X$, a quantity that measures the \emph{treeness} of $4$-tuples of points in $X$. This bound is known to be asymptotically tight. We improve this bound by restricting ourselves to metric spaces arising from filtered posets. By doing so we are able to replace the cardinality appearing in Gromov's bound by a certain poset theoretic invariant (the maximum length of fences in the poset) which can be much smaller thus significantly improving the approximation bound. The setting of metric spaces arising from posets is rich: For example, save for the possible addition of new vertices, every finite metric graph can be induced from a filtered poset. Since every finite metric space can be isometrically embedded into a finite metric graph, our ideas are applicable to finite metric spaces as well. At the core of our results lies the adaptation of the Reeb graph and Reeb tree constructions and the concept of hyperbolicity to the setting of posets, which we use to formulate and prove a tree approximation result for any filtered poset.

math.MG↗