Research

Here is a list of research projects I am involed in:

Shape Reconstruction

title: Shape Reconstruction publications: [“fw17”,“fasy2018reconstruction”] people: [ “Sushovan Majhi”,“Carola Wenk”, “Brittany Fasy”, “Rafal Komendarczyk”] permalink: research/shape-construction image: ‘/assets/projects/recon/graph-rips.jpg’

Application of algebraic topological methods in the reconstruction of geodesic spaces claims the lion’s share of my current research. The project is primarily motivated by the problem of reconstructing road-networks or maps from noisy GPS locations of drivers. Our goal is to find an efficient way to approximate a (hidden) map–both topologically and geometrically–from GPS locations sampled around it.

A road-network is often modeled as an embedded metric graph: a compact subset of $$\mathbb{R}^N$$ that is homeomorphic to a one-dimensional simplicial complex. We observe that the standard Euclidean metric $$\|\cdot\|$$ induces an intrinsic geodesic metric $$d_L$$ which, in this case, is the shortest-path distance on the graph. Such an observation naturally alluded to the geodesic subspaces on Euclidean space. We generalize the set-up and consider the reconstruction problem of geodesic subspaces of $$\mathbb{R}^N$$. We define a compact subset $$X$$ of $$\mathbb{R}^N$$ is to be a geodesic space if $$X$$ is a length metric space under the geodesic metric $$d_L$$. Our objective is to construct a topological space $$\widetilde{X}$$ from a (dense, but noisy) sample $$S$$ that is homotopy equivalent to the (unknown) geodesic space $$X$$. If such a strong reconstruction result is difficult to apprehend, we would like to at least compute the Betti numbers of $$X$$ from $$S$$.

In order to construct a topological space from a discrete set of sample points, the use of Vietoris-Rips and Čech complexes is becoming increasingly popular in TDA; see DeSilva:2004,ATTALI2013448,Adams_2019 for example. We investigate these complexes on the sample at different scales for a successful topological and geometric reconstruction.

The challenges in developing a geometrically close and computationally efficient reconstruction algorithm are ensued from the noise present in the data. We realize that the available topological methods for shape reconstruction only work for manifolds (SMALE) and a very special class of other compact Euclidean subsets: spaces with positive weak feature size (CHAZALSTAB). Because of the presence of corners and branching, these methods do not apply to our shapes of interest. We also observe in Figure 1 that the Vietoris-Rips (or Čech) complex on a dense sample around such a shape may not faithfully reflect its topology.

In fw17, we consider the special case of metric graphs and introduce the notion of geodesic feature size in order to develop an algorithm for computing the $$1$$-Betti number of the underlying of planar graph.

The 1-Betti of the underlying geodesic space (blue) is $$8$$. However, the Vietoris-Rips complex (red) on a dense sample (black) is shown to have its 1-Betti $$9$$. The picture is generated on smajhi.com/shape-reconstruction.

In order to further our understanding of the new feature size for a general geodesic space, we introduce two new sampling parameters: convexity radius and distortion of embedding for a geodesic space. In fasy2018reconstruction, we show when the point-cloud is sampled around a small Hausdorff proximity of the underlying geodesic space, both the Euclidean Čech and Vietoris-Rips complexes of the point-cloud can be used to correctly compute the homotopy (and homology) groups of the underlying shape. When considered the persistent topological features at small scales, we guarantee a correct topological reconstruction.

For a geometric reconstruction, in the special case of planar metric graphs, we further compute a Vietoris-Rips complex of the sample with respect to a different metric, which–unlike the Euclidean metric–embodies the underlying geodesic metric $$d_L$$ of the shape. Then, the shadow (as defined in Chambers2010) of this complex is shown to be homotopy equivalent to the ground truth.

As a coding hobbyist, I develop tools, libraries, and software to supplement my research. In order to demonstrate and visualize our theoretical development, I coded a JavaScript library, which is hosted on Github. The library runs on a web-app: smajhi.com/shape-reconstruction.

The success of our methodologies in fasy2018reconstruction emboldens our future objectives for a stronger reconstruction. So far, we successfully reconstruct the topology of general geodesic spaces. We also reconstruct the geometry of a planar embedded graph.

1. Our method does not currently extend to non-planar graph because of our lack of knowledge about the second homotopy group of the shadow. We conjecture that all its higher homotopy groups are trivial.

2. Currently, the output of such a geometric reconstruction is a thick region around the hidden graph. We are considering a post-processing step to prune the output shadow $$\widetilde{G}$$ in order to output an embedded graph that is isomorphic to the hidden graph $$G$$.

3. We also consider the geometric reconstruction of higher-dimensional simplicial complexes. Unlike the graphs, such a space may have non-trivial homotopy groups. We have seen that the output of computes the fundamental group correctly. However, it is not clear whether the higher homotopy groups of the output are also isomorphic to the respective homotopy groups of the underlying space.

Morse Theory

title: Graph Reconstruction via Discrete Morse Theory people: [“Sushovan Majhi”,“Carola Wenk”,“Brittany Fasy”] publications: [“fw18”] image: ‘/assets/projects/recon/graph-rips.jpg’

Gromov–Hausdorff Distance

title: Approximating Gromov-Hausdorff Distance people: [“Carola Wenk, Tulane University”, “Jeffrey Vitter, University of Mississippi”] publications: [“majhi2019approximating”] image: ‘/assets/projects/recon/graph-rips.jpg’

The Gromov-Hausdorff distance has been shown in memoli_theoretical_2005,memoli_use_nodate to provide a reasonable framework for comparing shapes. This metric was first introduced by M. Gromov in the context of Riemannian manifolds; see Gromov’s book gromov_metric_2007. However, it can also be defined between two abstract metric spaces $$(X,d_X)$$ and $$(Y,d_Y)$$ to measure how much they are metrically-close. Motivated by its potential use in topological data analysis, we address the problem of computing the Gromov-Hausdorff distance between two (finite) metric spaces.

Given two (compact) metric spaces $$(X,d_X)$$ and $$(Y,d_Y)$$, their Gromov-Hausdorff distance is formally defined (gromov_metric_2007) by

$d_{GH}(X,Y)=\inf_{\substack{f:X\to Z\\g:Y\to Z\$$Z,d_Z)}}d^Z_H(f(X),g(Y)),$ where the Hausdorff distance \(d^Z_H$$ of $$Z$$ is minimized over all isometries $$f,g$$ and metric spaces $$(Z,d_Z)$$.

Sometimes, an alternative definition, which is computationally more viable, is considered using the concept of correspondences. Given two point sets $$X$$ and $$Y$$, a correspondence $$C$$ between them is a subset of $$X\times Y$$ such that $$\forall x\in X\exists y\in Y\ni(x,y)\in C$$ and $$\forall y\in Y\exists x\in X\ni(x,y)\in C$$. Roughly speaking, a correspondence is a relation that is bi-surjective’’. Any surjective function $$f:X\to Y$$ naturally induces a correspondence

$C_f=\left\{\big(x,f(x)\big)\mid x\in X\right\}.$

For a given correspondence $$C$$, its distortion is defined to be

$dist(C)=\sup_{\substack{(x_1,y_1),(x_2,y_2)\in C}}|d_X(x_1,x_2)-d_Y(y_1,y_2)|.$

Now, the Gromov-Hausdorff distance is characterized by

$d_{GH}(X,Y)=\frac{1}{2}\inf_{C\subseteq X\times Y}dist(C);$

see burago_course_2001 for a proof. We consider the problem of efficiently computing the Gromov-Hausdorff distance between finite subsets $$X$$ and $$Y$$ of $$\mathbb{R}^d$$ equipped with the standard Euclidean metric.

We note from the latter definition of the Gromov-Hausdorff distance that it takes at least $$2^k$$ operations to naively compute it if $$X,Y$$ have at most $$k$$ points. In such a case in general, we seek an approximation algorithm that requires only polynomial many operations, but at the expense of computing an approximate answer that is within a certain approximation factor.

The authors of agarwal_computing_2015 show the NP-hardness of approximation of $$d_{GH}$$ when $$X,Y$$ are metric trees. However, computing the Gromov-Hausdorff distance between Euclidean subsets is still elusive.

In hutchison_approximating_2005, the authors consider (finite) subsets $$X,Y$$ of $$\mathbb{R}^d$$ of same cardinality and the class of only bijective correspondences between them. For $$d=1$$, they present an approximation algorithm to find the least distortion bijection with an approximation factor of $$2$$. They then pose two open problems: Can an approximation factor better than $$2$$ be achieved and is it NP-hard to compute the least distortion bijective correspondence between Euclidean subsets?

In our effort to approximate $$d_{GH}$$, we look into its relationship with $$d_{H,iso}$$, the minimum Hausdorff distance under the class of Euclidean isometries. For $$d=1$$ and compact subsets $$X,Y$$, we show in majhi2019approximating that

$d_{GH}(X,Y)\leq d_{H,iso}(X,Y)\leq\left(1+\frac{1}{4}\right)d_{GH}(X,Y).$

We also show that the upper bound is tight. This gives rise to an $$O(k\log{k})$$-time algorithm for approximating $$d_{GH}(X,Y)$$ with an approximation factor of $$\left(1+\frac{1}{4}\right)$$; see majhi2019approximating. Our method can also be adapted to have the same approximation factor for finding the lowest distortion bijective correspondences, hence solving the open problem addressed in hutchison_approximating_2005.

We now investigate the following two directions:

1. For $$d=1$$, we conjecture that it is NP-hard to approximate the Gromov-Hausdorff distance with a factor better than $$\left(1+\frac{1}{4}\right)$$.

2. Unlike $$d=1$$, the $$d_{H,iso}$$ fails to provide an approximation algorithm with a constant approximation factor when $$d>1$$. In higher dimension, are there any other ways to approximate the Gromov-Hausdorff distance?