
We propose a modification of the Mapper algorithm that improves its consistency for datasets whose local lens-space density varies greatly. This is motivated by temporal topic modelling, where lens-space density is temporal frequency and has high variance in many real-world datasets. For such datasets, tuning the resolution parameters of Mapper is difficult because small parameter changes can lead to significant changes in the output. Our modifications address this challenge by introducing kerneled covers that permit Mapper’s cover resolution to change locally in semantic space, and then using these covers to have the resolution adapt to the local lens-space density of the dataset. The core mechanism of our proposed covers is that they are determined by a kernel function whose width determines the cover resolution and increases in regions with lower lens-space density. We prove that Mapper with kerneled covers converges in bottleneck distance to the Reeb graph of the Rips complex of the data, generalizing the analogous result for Mapper. Moreover, we discuss implementation details and how one can apply Mapper to temporal topic modelling and provide the results of computational experiments, including a temporal topic model of a real-world dataset. There is an accompanying reference implementation.
We investigate how effectively finite metric spaces can be distinguished by distance-based invariants. As model spaces, we consider regular polygons, cycle graphs, and their generalization, circular metric spaces, and as invariants we consider the distance multiset, magnitude, and magnitude homology. We construct explicit families of homometric but non-congruent circular metric spaces, and in many even cases these examples also have the same magnitude as the original space. We prove that regular polygons are determined by the distance multiset among planar metric spaces, but not in general. We also determine, for several values of n, whether regular n-gons and n-cycle graphs are determined by magnitude.
The distance matrix of a finite point cloud can be visualized as a heatmap. When the data arise from a time series, the sublevel sets of this image are known as recurrence plots and are widely used in time series analysis. Motivated by this perspective, we establish a relationship between the distance matrix filtration of the time series and the Čech (or Vietoris–Rips) filtration of its state space embedding in the form of a degree-one chain map. We study the induced maps in homology, showing that the map from H_0 into H_1 is surjective and providing an example where the map from H_1 into H_2 is nonzero. These chain maps can be applied to simplify image persistence computations arising in the computation of cycling signatures, a topological tool for time series analysis. Moreover, these computations yield finer information that allows the analysis of transitions between different types of cycling motion.
In this paper we study parametrized topological complexity of bundles of real projective spaces which arise as projectivisations of vector bundles. We develop algebraic machinery for computing lower bounds for the parametrized topological complexity of projective bundles based on theory of Stiefel - Whitney characteristic classes. We establish sharp upper bounds for the parametrized topological complexity of projective bundles improving the general upper bounds. Combining the lower and the upper bounds we compute explicitly many specific examples.
In this paper, we study a new construction which associates a combinatorial cubical complex (G) to an arbitrary undirected simple graph G. The vertices of (G) are indexed by all possible orientations of the edges of G. The cells of (G) are the sets of independent flexes, where a flex is a simultaneous change of orientations of the edges adjacent to a certain sink or a certain source in G. Accordingly, we call (G) the flex complex of the graph G. Our focus is on studying topology and combinatorics of the flex complexes. The main topological theorem says that for an arbitrary graph G, the flex complex (G) is homotopy equivalent to a disjoint union of tori. We also provide formulae for the number of these tori. Furthermore, we prove a much more precise combinatorial result saying that when G is connected, every connected component of (G) is either a collapsible cubical complex, or can be collapsed to a cycle whose length is equal to the number of vertices of G. We shall provide a combinatorial enumeration for the components of both types. Our study is motivated by the beauty and naturality of the graph construction, as well as by the mathematical modeling of the network evolution.