Abstracts

Second Mexican School/Conference on

Topological Data Analysis and Related Topics
Juriquilla, Queretaro, Mexico
December 7-11, 2015

Ulrich Bauer, Technical University of Munich, Germany.

Title. Induced matchings and the algebraic stability of persistence barcodes.
Monday, December 7th, 16.3017.20.
Abstract. We define a simple, explicit map sending a morphism f : M N of pointwise
finite dimensional persistence modules to a matching between the barcodes of M and N. Our
main result is that, in a precise sense, the quality of this matching is tightly controlled by the
lengths of the longest intervals in the barcodes of ker(f ) and coker(f ).
As an immediate corollary, we obtain a new proof of the algebraic stability theorem for
persistence barcodes, a fundamental result in the theory of persistent homology. In contrast to
previous proofs, ours shows explicitly how a interleaving morphism between two persistence
modules induces a matching between the barcodes of the two modules. Our main result also
specializes to a structure theorem for submodules and quotients of persistence modules, and
yields a novel single-morphism characterization of the interleaving relation on persistence
modules.

Peter Bubenik, University of Florida, USA.

Title. Statistical topological data analysis.
Tuesday, December 8th, 09.0009.50.
Abstract. Topological data analysis (TDA) uses topological methods to summarize the
shape of data. I will introduce TDA and some of its standard constructions. In this context,
it is natural to ask certain statistical questions. Some of these are hard to address using the
traditional constructions. I will present an alternative summary, the persistence landscape,
describe some its properties, and show how it may be used for hypothesis testing. I will apply
these ideas to study protein data.

Peter Bubenik, University of Florida, USA.

Title. Topological data analysis and machine learning.
Wednesday, December 9th, 09.0009.50.
Abstract. Topological data analysis (TDA) provides summaries of the shape of data. If
these summaries are chosen to be vectors, then one can easily apply tools from statistics and
machine learning to the output of TDA. I will discuss a number of ways to obtain such vector
summaries using TDA. I will show how one such summary, the persistence landscape, may be
combined with statistics and machine learning to study brain artery and protein data.

Peter Bubenik, University of Florida, USA.

Title. Generalized persistence.
Thursday, December 10th, 09.0009.50.
Abstract. I will introduce a formal framework in which one can consider persistent homology, the central tool of topological data analysis. This formalism can be helpful for understanding certain algebraic relations and constructions for persistence: interleavings and
interpolations. It also provides a setting for generalizations of persistent homology.

Gunnar Carlsson, Ayasdi and Stanford University, USA.

Title. Topology and the Big Data problem.
Monday, December 7th, 09.0009.50.
Abstract. There has been discussion of Big Data. Although we are able to collect data
more and more efficiently, the major roadblock toward making data useful is in developing
analysis techniques for it. One approach to this problem is in an organization principle involving
the shape of the data, i.e. the imposition of a metric on it. Since topology is the part of
mathematics which studies shape, it is natural to attempt to translate techniques from this
subject into the world of finite metric spaces. We will discuss these methods, with numerous
examples.

Armando Casta
neda, Universidad Nacional Autonoma de Mexico, Mexico.
Title. Computing independent sets in an asynchronous environment.
Friday, December 11th, 11.3012.20.
Abstract. We consider an independent set task for a set of asynchronous crashprone
processes that have to output an independent set of a graph G, as large as possible. Processes
communicate through a read/write shared-memory. Each process starts with a preference for a
vertex in G, communicates with other processes, and decides on an output vertex. The output

vertices are distinct, and no two output vertices belong to the same edge. Furthermore, if all
initial preferences are distinct, and form an independent set of G, then the output vertex of
each process is equal to its initial preference. The independent set number of G is the largest
number of processes n that can solve the task on G.
We propose a waitfree distributed independent set algorithm (tolerating any number of
process crashes), and show that it has optimal independent set number, among static independent set algorithms. Also, we prove that for many graphs, no independent set algorithm can
do better. In a static independent set algorithm there is an independent set I such that in
every execution, if a process does not decide its initial input, it decides a vertex in I. We use
topological techniques to prove our impossibility results.

Frederic Chazal, INRIA Saclay, France.

Title. Subsampling methods and Wasserstein stability for persistent homology.
Wednesday, December 9th, 16.3017.20.
Abstract. Computational topology has recently seen an important development toward
data analysis, giving birth to Topological Data Analysis. Persistent homology appears as a
fundamental tool in this field. It is usually computed from filtrations built on top of data sets
sampled from some unknown (metric) space, providing topological signatures revealing the
structure of the underlying space. When the size of the sample is large, direct computation of
persistent homology often suffers two issues. First, it becomes prohibitive due to the combinatorial size of the considered filtrations and, second, it appears to be very sensitive to noise and
outliers.
In this talk we will present a method to overcome these issues by computing persistent
diagrams from several subsamples and combining them in order to efficiently infer robust and
relevant topological information from data.

Herbert Edelsbrunner, Institute of Science and Technology, Austria.

Title. Alpha shapes and incremental Betti numbers.
Monday, December 7th, 10.0010.50.
Abstract. In the early 80s of last century, we invented alpha shapes as a reaction to the
quest of defining the shape of a finite point set in a way that is more adaptive than the convex
hull. The implementation of this idea in 3 dimensions was complicated and required advances
in writing geometric software. We succeeded in the early 90s. Having the filtration of alpha
shapes available, we developed an incremental algorithm to add homology information at all
scales.
The main steps in this research were conducted in collaboration with David Kirkpatrick,
Raimund Seidel, Ernst-Peter Muecke, Harald Rosenberger, and Cecil Delfinado.

Herbert Edelsbrunner, Institute of Science and Technology, Austria.

Title. Persistent homology and two algorithms.
Tuesday, December 8th, 10.0010.50.
Abstract. Observing the practical shortcomings of the Betti numbers for protein data,
we invented the concept of persistence to add a quantitative component to the existence of
holes. Extending the incremental Betti number algorithm, we added a search procedure to
find the earliest cycles destroyed by a new simplex. This turned out to be a sparse-matrix
implementation of a particular matrix reduction method.
The main steps in this research were conducted in collaboration with David Letscher and
Afra Zomorodian.

Herbert Edelsbrunner, Institute of Science and Technology, Austria.

Title. Stability and intrinsic volumes.
Wednesday, December 9th, 10.0010.50.
Abstract. The most useful property of persistence is the stability of its diagram. Proving
it for general functions was complicated, but there is a more straightforward argument for the
case of two PL functions defined on the same triangulation. To illustrate the centrality of this
theorem, we exhibit a connection to intrinsic volumes. Defining them with Croftons integral
geometry formula, we get continuity for convex but not for non-convex bodies. Using persistent
homology we rig the definition with the goal to get continuity also for non-convex shapes.
The main steps in this research were conducted in collaboration with David Cohen-Steiner,
John Harer, Dmitriy Morozov, and Florian Pausinger.

Lisbeth Fajstrup, Aalborg University, Denmark.

Title. Directed topologyConcurrency theory.
Friday, December 11th, 12.3013.20.
Abstract. Structural topological insights become more and more important in the discovery
and analysis of fundamental mechanisms in science and engineering. In this survey talk, I want
to report on a specific recent such development arising from models for concurrency in Computer
Science.
Directed algebraic topology, topological spaces with a time direction, was introduced almost 20 years ago to study new models of concurrent computing Higher Dimensional Automata and PV-models. In these models, executions are timedirected paths and they are equivalent if they can be continuously deformed into each other, i.e., they are homotopic through
directed paths. Hence topology is the right tool, at least when the time direction can be taken
into account, and this is what directed topology does. This talk will give an introduction to the
mathematical field which is still growing and to the applications in concurrency, where both

new algorithms and insight have come from this point of view. There are still open problems
both on the mathematical side and on the specific model of concurrency.

Michael Farber, Queen Mary University of London, UK.

Title. Topology of large random spaces.
Monday, December 7th Wednesday, December 9th, 11.3012.20.
Abstract. Topological spaces (such as manifolds and simplicial complexes) are traditionally used in applications as models of mechanical systems. However, when dealing with large
systems the standard notion of a space becomes inadequate and for modelling such systems
one needs mathematical objects of a new kind. In my lectures I will describe several models producing large random simplicial complexes and recent results about the topological and
geometric properties of such spaces.

Patrizio Frosini, University of Bologna, Italy.

Title. An observer-oriented approach to topological data analysis - Part 1: From comparing
subsets of Rn to studying metric spaces of functions.
Thursday, December 10th, 10.0010.50.
Abstract. In many experiments concerning shape comparison, data can be represented by
functions defined on topological spaces. For example, if we wish to compare two images, we
can consider two functions from the real plane to R3 , where each triple of numbers describes
the color of a given point in the image. This observation naturally leads to an approach that
translates the comparison of data into the comparison of functions defined from a topological
space X to Rk .
An important element in this procedure is the concept of invariance. Indeed, two items in a
dataset are often judged equivalent if they are obtained by applying a transformation belonging
to a selected group G. For example, two photographs are usually considered equal if they are
obtained from each other by applying a rotation. In this case our quantitative comparison
should use a pseudo-metric that is invariant under the action of the group G, such as the
natural pseudo-distance dG associated with a group G of self-homeomorphisms of a topological
space X. If is the collection of functions from X to Rk that represent our data, the previous
pseudo-metric is defined by setting dG (1 , 2 ) = inf gG k1 2 gk for every 1 , 2 . It
is easy to prove that dG (1 g, 2 ) = dG (1 , 2 g) = dG (1 , 2 ) for every g G and every
1 , 2 .
In some sense, the approach based on the natural pseudo-distance dG disregards the objects producing the signals available to the observer and focuses on the functions describing
those signals, following the assumption that reality can be studied only by considering the measurements that we can make on it. In other words, we are not interested in a family {Si Rn }

of sets representing objects, but in a metric space of functions that describe experimental
measurements and are defined on a single topological space X.
Unfortunately, the direct study of dG is difficult and computationally expensive, because the
group G is usually quite large. However, some techniques are available to obtain lower bounds
for dG . In particular, in a recent paper we have proven that good lower bounds can be obtained
by applying G-invariant non-expansive operators (GINOs) and persistent homology to the
functions in , for any choice of the group G. These GINOs can be seen as the actions made by
the observer on the data before analyzing them, and reflect her/his invariance criteria. Filtering
and averaging are examples of such actions. In this talk we will illustrate this mathematical
setting, together with some theoretical results.
This research has been conducted in collaboration with Grzegorz Jablo
nski (Jagiellonian
University, Krakow, Poland).

Patrizio Frosini, University of Bologna, Italy.

Title. An observer-oriented approach to topological data analysis - Part 2: The algebra of
group invariant non-expansive operators and its application in the project GIPHOD Group
Invariant Persistent Homology Online Demonstrator.
Friday, December 11th, 10.0010.50.
Abstract. Following the ideas presented in Part 1, in this talk we will illustrate the algebra
of group invariant non-expansive operators and some experimental results showing how this
algebra can be used to compare simple synthetic grey-level images with respect to some given
invariance groups. These groups can be seen as invariance criteria chosen by an observer. Our
results have been produced by means of the forthcoming online demonstrator GIPHOD, which
is in development to present the use of G-invariant persistent homology for image comparison.
This research has been conducted in collaboration with Grzegorz Jablo
nski (Jagiellonian
University, Krakow, Poland) and Marc Ethier (Universite de Saint-Boniface, Winnipeg, Manitoba, Canada).

Heather Harrington, University of Oxford, UK.

Title. Topological data analysis for investigation of dynamics and networks.
Thursday, December 10th, 16.3017.20.
Abstract. Persistent homology (PH) is a technique in topological data analysis that allows
one to examine features in data across multiple scales in a robust and mathematically principled
manner, and it is being applied to an increasingly diverse set of applications. We investigate
applications of PH to dynamics and networks, focusing on two settings: dynamics on a network
and dynamics of a network.
We analyze a contagion spreading on a network using persistent homology. Next we investigate a network that changes in time and show that persistent homology may be useful for

distinguishing temporal distributions and a high level summary of temporal structure. Together, these two investigations illustrate that persistent homology can be very illuminating in
the study of networks and their applications.

Wolfgang Heil, Florida State University, USA.

Title. Simply connected 2-stratifolds.
Tuesday, December 8th, 16.3017.20.
Abstract. 2-stratifolds are the simplest 2-complexes which can be good models for topological data analysis. A 2-stratifold X contains a collection X 1 of finitely many simple closed
curves such that X X 1 is a 2-manifold and a neighborhood of each component C of X 1 consists of sheets intersecting in C. In contrast to 2-manifolds there is no known classification of
2-stratifolds in terms of algebraic invariants. In this talk we will describe 2-stratifolds and their
graphs and present efficient algorithms that detect whether certain 2-stratifolds have trivial
first homology, whether they are simply connected, or whether they have the same homotopy
type as a 2-sphere.
This is joint work with J. C. Gomez-Larra
naga and F. Gonzalez-Acu
na.

Maurice Herlihy, Brown University, USA.

Title. Distributed computing through combinatorial topology.
Monday, December 7th Thursday, December 10th, 15.3016.20.
Sergio Rajsbaum, Universidad Nacional Autonoma de Mexico, Mexico.
Introductory Lecture. An overview of distributed computing through combinatorial topology.
Monday, December 7th, 12.3013.20.
Abstract. In the past two decades, exciting new topological techniques have emerged for
analyzing distributed algorithms. This research has resulted in dozens of articles, and the
awarding of the 2004 Godel prize (the highest prize in theoretical computer science), to the
researchers that uncovered the intimate connection between topology and distributed computing. These techniques have enabled the development of a solid theoretical foundation relevant
to many real systems where collaboration via parallelism with unpredictable delays is essential, such as multicore microprocessors, wireless networks, distributed systems, and Internet
protocols.
An overview of this area will be presented, followed by a mini-course based on the recent book
Distributed Computing Through Combinatorial Topology by Herlihy, Kozlov, Rajsbaum,
Elsevier.

Neza Mramor Kosta, University of Ljubljana, Slovenia.

Title. Discrete Morse theory with applications to data.
Tuesday, December 8th Thursday, December 10th, 12.3013.20.
Abstract. In the last two decades since it was introduced, discrete Morse theory has proven
to be an extremely useful tool in topological data analysis. We will first give an overview of the
theory and its main results. Examples of applications to data from different domains, demonstrating how discrete Morse theoretical techniques can be used to extract useful information
from data, as well as aid in solving problems like classification, clustering and prediction, will
be presented.

Nina Otter, University of Oxford, UK.

Title. A roadmap for the computation of persistent homology.
Friday, December 11th, 09.0009.50.
Abstract. Persistent homology is a method from algebraic topology used in topological
data analysis to study qualitative features of data. It is robust to small perturbations in
measurement, independent of dimension and scales and provides a compact representation of
the outputs. Despite recent progress, the computation of persistent homology remains an open
area of research with numerous important and fascinating challenges. It is a rapidly evolving
field, with new algorithms and implementations being developed and released at a fast pace.
In this talk I will give an overview of the state-of-the-art algorithms and software for the
computation of persistent homology, outline the challenges related to the computation of persistent homology, and indicate some points that I believe should be addressed by the TDA
community to overcome the current limitations.
This talk is based on joint work with M. Porter, U. Tillmann, P. Grindrod and H. Harrington.

Antonio Rieser, CONACYT-CIMAT, Mexico.

Title. Homotopy theory for data sets.
Thursday, December 10th, 11.3012.20.
Abstract. A nearly ubiquitous assumption in modern data analysis is that a given highdimensional data set concentrates around a lower dimensional space. Recently, a great deal
of attention has been focused on how to use point samples from a metric measure space to
estimate the topological and geometric invariants of the space, and on applying the resulting
algorithms to real data sets. Most techniques for studying the topology of data, in particular
persistent homology, proceed by considering families of topological spaces which were in some
way thicker than the original set. In this talk, we ask instead what sort of homotopy theory
one may construct directly on finite sets of points. We construct such a homotopy theory, show
how it may be applied to a variety of combinatorial objects, and give several computations for
point clouds, graphs, and simplicial complexes.