In Image Classification, pictures of the same type of object can have very different pixel values. Traditional norm-based metrics therefore fail to identify objectsin the same category. Topology is a branch of mathematics that deals with homeomorphic spaces, by discarding length. With topology, we can discover patterns in the image that are invariant to rotation, translation and warping. Persistent Homology is a new approach in Applied Topology that studies the presence of continuous regions and holes in an image. It has been used successfully for image segmentation and classification [12]. However, current approaches in image classification require a grayscale image to generate the persistence modules. This means information encoded in colour channels is lost. This thesis investigates whether the information in the red, green and blue colour channels of an RGB image hold additional information that could help algorithms classify pictures. We apply two recent methods, one by Adams [2] and the other by Hofer [25], on the CUB-200-2011 birds dataset [40] andfind that Hofer’s method produces significant results. Additionally, a modified method based on Hofer that uses the RGB colour channels produces significantly better results than the baseline, with over 48 % of images correctly classified, compared to 44 % and with a more significant improvement at lower resolutions.This indicates that colour channels do provide significant new information and generating one persistence module per colour channel is a viable approach to RGB image classification.

Topological Data Analysis is a quickly expanding field but one particular subfield, multidimensional persistence, has hit a dead end. Although TDA is a very active field, it has been proven that the one-dimensional persistence used in persistent homology cannot be generalized to higher dimensions. With this in mind, progress can still be made in the accuracy of approximating it. The central challenge lies in the multiple persistence parameters. Using more than one parameter at a time creates a multi-filtration of the data which cannot be totally ordered in the way that a single filtration can. The goal of this thesis is to contribute to the development of persistence heat maps by replacing the persistent betti number function (PBN) defined by Xia and Wei in 2015 with a new persistence summary function, the accumulated persistence function (APF) defined by Biscio and Moller in 2016. The PBN function fails to capture persistence in most cases and thus their heat maps lack important information. The APF, on the other hand, does capture persistence that can be seen in their heat maps. A heat map is a way to visually describe three dimensions with two spatial dimensions and color. In two-dimensional persistence heat maps, the two chosen parameters lie on the x- and y- axes. These persistence parameters define a complex on the data, and its topology is represented by the color. We use the method of heat maps introduced by Xia and Wei. We acquired an R script from Matthew Pietrosanu to generate our own heat maps with the second parameter being curvature threshold. We also use the accumulated persistence function introduced by Biscio and Moller, who provided an R script to compute the APF on a data set. We then wrote new code, building from the existing codes, to create a modified heat map. In all the examples in this thesis, we show both the old PBN and the new APF heat maps to illustrate their differences and similarities. We study the two-dimensional heat maps with respect to curvature applied to two types of parameterized knots, Lissajous knots and torus knots. We also show how both heat maps can be used to compare and contrast data sets. This research is important because the persistence heat map acts as a guide for finding topologically significant features as the data changes with respect to two parameters. Improving the accuracy of the heat map ultimately improves the efficiency of data analysis. Two-dimensional persistence has practical applications in analyses of data coming from proteins and DNA. The unfolding of proteins offers a second parameter of configuration over time, while tangled DNA may have a second parameter of curvature. The concluding argument of this thesis is that using the accumulated persistence function in conjunction with the persistent betti number function provides a more accurate representation of two-dimensional persistence than the PBN heat map alone.

<p>This dissertation extends the theory of persistent homology to time varying systems. Most of the previous work has been dedicated to using this powerful tool in topological data analysis to study static point clouds. In particular, given a point cloud, we can construct its persistence diagram. Since the diagram varies continuously as the point cloud varies continuously, we study the space of time varying persistence diagrams, called vineyards when they were introduced by Cohen-Steiner, Edelsbrunner, and Morozov.</p><p>We will first show that with a good choice of metric, these vineyards are stable for small perturbations of their associated point clouds. We will also define a new mean for a set of persistence diagrams based on the work of Mileyko et al. which, unlike the previously defined mean, is continuous for geodesic vineyards. </p><p>Next, we study the sensor network problem posed by Ghrist and de Silva, and their application of persistent homology to understand when a set of sensors covers a given region. Giving each of these sensors a probability of failure over time, we show that an exact computation of the probability of failure of the whole system is NP-hard, but give an algorithm which can predict failure in the case of a monitored system.</p><p>Finally, we apply these methods to an automated system which can cluster agents moving in aerial images by their behaviors. We build a data structure for storing and querying the information in real-time, and define behavior vectors which quantify behaviors of interest. This clustering by behavior can be used to find groups of interest, for which we can also quantify behaviors in order to determine whether the group is working together to achieve a common goal, and we speculate that this work can be extended to improving tracking algorithms as well as behavioral predictors.</p> / Dissertation

<p>In this thesis we explore and extend the theory of persistent homology, which captures topological features of a function by pairing its critical values. The result is represented by a collection of points in the extended plane called persistence diagram.</p><p>We start with the question of ridding the function of topological noise as suggested by its persistence diagram. We give an algorithm for hierarchically finding such epsilon-simplifications on 2-manifolds as well as answer the question of when it is impossible to simplify a function in higher dimensions.</p><p>We continue by examining time-varying functions. The original algorithm computes the persistence pairing from an ordering of the simplices in a triangulation and takes worst-case time cubic in the number of simplices. We describe how to maintain the pairing in linear time per transposition of consecutive simplices. A side effect of the update algorithm is an elementary proof of the stability of persistence diagrams. We introduce a parametrized family of persistence diagrams called persistence vineyards and illustrate the concept with a vineyard describing a folding of a small peptide. We also base a simple algorithm to compute the rank invariant of a collection of functions on the update procedure.</p><p>Guided by the desire to reconstruct stratified spaces from noisy samples, we use the vineyard of the distance function restricted to a 1-parameter family of neighborhoods of a point to assess the local homology of a sampled stratified space at that point. We prove the correctness of this assessment under the assumption of a sufficiently dense sample. We also give an algorithm that constructs the vineyard and makes the local assessment in time at most cubic in the size of the Delaunay triangulation of the point sample.</p><p>Finally, to refine the measurement of local homology the thesis extends the notion of persistent homology to sequences of kernels, images, and cokernels of maps induced by inclusions in a filtration of pairs of spaces. Specifically, we note that persistence in this context is well defined, we prove that the persistence diagrams are stable, and we explain how to compute them. Additionally, we use image persistence to cope with functions on noisy domains.</p> / Dissertation

Brain decoding involves the determination of a subject’s cognitive state or an associated stimulus from functional neuroimaging data measuring brain activity. In this setting the cognitive state is typically characterized by an element of a finite set, and the neuroimaging data comprise voluminous amounts of spatiotemporal data measuring some aspect of the neural signal. The associated statistical problem is one of classification from high-dimensional data. We explore the use of functional principal component analysis, mutual information networks, and persistent homology for examining the data through exploratory analysis and for constructing features characterizing the neural signal for brain decoding. We review each approach from this perspective, and we incorporate the features into a classifier based on symmetric multinomial logistic regression with elastic net regularization. The approaches are illustrated in an application where the task is to infer from brain activity measured with magnetoencephalography (MEG) the type of video stimulus shown to a subject. / Graduate

abstract: Topological methods for data analysis present opportunities for enforcing certain invariances of broad interest in computer vision: including view-point in activity analysis, articulation in shape analysis, and measurement invariance in non-linear dynamical modeling. The increasing success of these methods is attributed to the complementary information that topology provides, as well as availability of tools for computing topological summaries such as persistence diagrams. However, persistence diagrams are multi-sets of points and hence it is not straightforward to fuse them with features used for contemporary machine learning tools like deep-nets. In this paper theoretically well-grounded approaches to develop novel perturbation robust topological representations are presented, with the long-term view of making them amenable to fusion with contemporary learning architectures. The proposed representation lives on a Grassmann manifold and hence can be efficiently used in machine learning pipelines. The proposed representation.The efficacy of the proposed descriptor was explored on three applications: view-invariant activity analysis, 3D shape analysis, and non-linear dynamical modeling. Favorable results in both high-level recognition performance and improved performance in reduction of time-complexity when compared to other baseline methods are obtained. / Dissertation/Thesis / Masters Thesis Electrical Engineering 2017

La théorie de l'homologie généralise en dimensions supérieures la notion de connectivité dans les graphes. Étant donné un domaine, décrit par un complexe simplicial, elle définit une famille de groupes qui capturent le nombre de composantes connexes, le nombre de trous, le nombre de cavités et le nombre de motifs équivalents en dimensions supérieures. En pratique, l'homologie permet d'analyser des systèmes de données complexes, interprétés comme des nuages de points dans des espaces métriques. La théorie de l'homologie persistante introduit une notion robuste d'homologie pour l'inférence topologique. Son champ d'application est vaste, et comprend notamment la description d'espaces des configurations de systèmes dynamiques complexes, la classification de formes soumises à des déformations et l'apprentissage en imagerie médicale. Dans cette thèse, nous étudions les ramifications algorithmiques de l'homologie persistante. En premier lieu, nous introduisons l'arbre des simplexes, une structure de données efficace pour construire et manipuler des complexes simpliciaux de grandes dimensions. Nous présentons ensuite une implémentation rapide de l'algorithme de cohomologie persistante à l'aide d'une matrice d'annotations compressée. Nous raffinons également l'inférence de topologie en décrivant une notion de torsion en homologie persistante, et nous introduisons la méthode de reconstruction modulaire pour son calcul. Enfin, nous présentons un algorithme de calcul de l'homologie persistante zigzag, qui est une généralisation algébrique de la persistance. Pour cet algorithme, nous introduisons de nouveaux théorèmes de transformations locales en théorie des représentations de carquois, appelés principes du diamant. Ces algorithmes sont tous implémentés dans la librairie de calcul Gudhi. / The theory of homology generalizes the notion of connectivity in graphs to higher dimensions. It defines a family of groups on a domain, described discretely by a simplicial complex that captures the connected components, the holes, the cavities and higher-dimensional equivalents. In practice, the generality and flexibility of homology allows the analysis of complex data, interpreted as point clouds in metric spaces. The theory of persistent homology introduces a robust notion of homology for topology inference. Its applications are various and range from the description of high dimensional configuration spaces of complex dynamical systems, classification of shapes under deformations and learning in medical imaging. In this thesis, we explore the algorithmic ramifications of persistent homology. We first introduce the simplex tree, an efficient data structure to construct and maintain high dimensional simplicial complexes. We then present a fast implementation of persistent cohomology via the compressed annotation matrix data structure. We also refine the computation of persistence by describing ideas of homological torsion in this framework, and introduce the modular reconstruction method for computation. Finally, we present an algorithm to compute zigzag persistent homology, an algebraic generalization of persistence. To do so, we introduce new local transformation theorems in quiver representation theory, called diamond principles. All algorithms are implemented in the computational library Gudhi.

No description available.

Studying lymphocyte development using mass cytometry Abstract Development of mature lymphocytes, a white blood cell subtype, is crucial for the correct function of the human immune system. Currently, developmental pathways of lymphocytes can be studied using high-throughput single-cell measurements. In particular, mass cytometry enables the study of immunologically relevant pheno- typic and functional markers on a vast scale. In this work I present my individual contribution to tviblindi, a powerful software tool for analysis of cytometric data aimed at uncovering developmental trajectories. tviblindi is a package written in R, Python and C++. It provides a means to integrate prior knowledge with data analyses grounded in graph theory and algebraic topology. tviblindi is accessible to biological researchers without background in computer science or mathematics. It is an addition to the expanding field of trajectory inference in single-cell data. Furthermore, I review current knowledge of T-cell development and conduct a tviblindi analysis thereof using human thymus and peripheral blood datasets and evaluate the results. 1

We propose a model that generates a family network based on real-world family network data. We use this model to study the extent to which distances to union and the number of children characterize family networks. To determine how accurate our model is we use persistent homology to identify and compare the structure of our modeled family networks to real-world family networks. To accomplish this, we introduce the notion of a network's persistence curve, which encodes the network's set of persistence intervals. Using the bottleneck distance allows us to measure the difference in the homological structure between any pair of networks. We also study how the distribution of distance to union and the distribution of children build family networks. What we find is that these two features of distance to union and number of children allow us to fairly accurately recreate family networks at least at the level of their persistent homology.