PUB - Publikationen an der Universität Bielefeld

In this paper we present our reimplementation of the Growing Neural Gas Sonificationfor interactive Interfaces such as our t-Desk or touch-capable tablet PCs. Growing Neural Gas (GNG) is an undirected learning algorithm that incrementally ’grows’ a network graph into data distributions, revealing the data distributions intrinsic dimensionality and aspects of its structure. The GNG Sonification (GNGS) provides a method to interactively explore the GNG during the growing process, utilizing a Model-Based Sonification(MBS) to convey audible information about the data distribution in addition to the visualization. The goal of our reimplementation was to be able to rapidly grasp the structure of the sonified and visualized data, to give the user the ability to conduct direct A/B comparisons between different (or similar) clusters within a data distribution. The direct bi-manual interaction as well as a simplified fullscreen touchable user interface helps to focus on the exploration of the GNG rather than the interaction itself. We present and discuss different interaction metaphors for the excitation of the model setup in this MBS.