Uncertainty-Aware Multidimensional Scaling



David Hägele, Tim Krake, Daniel Weiskopf

 View presentation:2022-10-18T16:30:00ZGMT-0600Change your timezone on the schedule page
2022-10-18T16:30:00Z

Exemplar figure, described by caption below — Uncertainty-Aware Multidimensional Scaling applied to a set of 6-dimensional distributions. The distributions describe the variance in stats (Hit Points, Attack, Defense, Sp. Att., Sp. Def., Speed) for different Pokemon. The plot shows the projected probability densities with isolines for different quantiles.

Prerecorded Talk

The live footage of the talk, including the Q&A, can be viewed on the session page, VIS Opening (10:45am-11:03am)| Best Papers (11:03am-12:00pm).

Fast forward

Abstract

We present an extension of multidimensional scaling (MDS) to uncertain data, facilitating uncertainty visualization of multidimensional data. Our approach uses local projection operators that map high-dimensional random vectors to low-dimensional space to formulate a generalized stress. In this way, our generic model supports arbitrary distributions and various stress types. We use our uncertainty-aware multidimensional scaling (UAMDS) concept to derive a formulation for the case of normally distributed random vectors and a squared stress. The resulting minimization problem is numerically solved via gradient descent. We complement UAMDS by additional visualization techniques that address the sensitivity and trustworthiness of dimensionality reduction under uncertainty. With several examples, we demonstrate the usefulness of our approach and the importance of uncertainty-aware techniques.