Open Conference Systems, CLADAG2023

Font Size: 
New tour methods for visualizing high-dimensional data
Ursula Laa, Dianne Cook

Last modified: 2023-06-02

Abstract


Tour methods visualize high-dimensional spaces as animated sequences of low-dimensional projections. Viewing the projected data allows us to uncover and understand shapes and patterns in such high-dimensional spaces. Typically we create this animation by first selecting a target plane, and then we interpolate to gradually move to the selected target. While several methods for target selection are available in the R package tourr, it currently only implements geodesic interpolation. Here we present recent developments in tour methods. We first describe a manual user-guided control for target selection, that also includes the interactive selection of sections in the context of a slice tour. We then present a new interpolation method for frame-to-frame transitions instead of plane-to-plane, important for projection pursuit applications.