ClassMat: a Matrix of Small Multiples to Analyze the Topology of Multiclass Multidimensional Data

Michael Aupetit, Ahmed Ali, Abdelkader Baggag, Halima Bensmail

View presentation:2022-10-17T21:20:00ZGMT-0600Change your timezone on the schedule page
2022-10-17T21:20:00Z
Exemplar figure, described by caption below
ClassMat aims to help the analyst explore the topology of multidimensional labeled data by focusing their attention on within-class or between-class topology, one or two classes at a time. ClassMat organizes data visualizations in a matrix: one class per chart on-diagonal, and pairs of classes off-diagonal such that all pairs are uniquely represented. Chart dimensions are determined by the displayed classes: in this picture, Linear Discriminant projections are used off-diagonal, and Principal Components on-diagonal. ClassMat complements SPLOM/pairplot/GPLOM organizing the matrix per class labels rather than per dimensions of the data. As GPLOM, it extends beyond scatterplots.

The live footage of the talk, including the Q&A, can be viewed on the session page, TopoInVis: Session 2, Early Career Lightning Talks + Best Paper Awards .

Keywords

Visualization; topological data analysis;

Abstract

Data scientists often deal with multiclass multidimensional data. There is a need to support the exploration of such data with topological methods. We propose a new visualization metaphor for multiclass data and illustrate it with two complementary analytic approaches. We design ClassMat, a visualization matrix similar in spirit to the scatterplot matrix (SPLOM or pairs plot) but focused on pairs of classes rather than pairs of dimensions. We show how this visualization matrix can be used in two main multidimensional data visualization pipelines: visualization-then-topological-analysis and topological-analysis-then-visualization. In particular, we show it can support the analyst in detecting interferences between topological features of different classes.