PUB - Publikationen an der Universität Bielefeld

ŒThis paper presents a novel interactive auditory data exploration method to investigate features of high-dimensional data distributions. ŒThe Mode Explorer couples a scratching-interaction on a 2D scatterŠ plot of high-dimensional data to real-time dynamical processes, excited in data space at the nearest mode in the probability density function (pdf) obtained by kernel-density estimation. Speci€cally, the sign-inverted pdf is used as a potential function in which test particles perform oscillations at low friction, yielding signals that can directly be played back as sound. ThŒis Model-based soni€fication approach is used to interactively search the distribution for diff‚erent modes, learn about their details, i.e. the Hessian matrix at the mode, and thus enable a non-parametric parameter selection for appropriate bandwidth. Files: V1.mov: Demonstration of sonifying the particles as they traveled through the data space and converged to the mode with a low friction. V2.mov: Demonstration video of the Mode Explorer using a Wacom Tablet and modified stylus. trajectory.txt: Cython code of the potential function. Note: This code only demonstrates the algorithm, to use it Cython library needs to be installed.