Abstract
The combination of pixelization and dimensional stacking uniquely facilitates the visualization and analysis of large, multidimensional databases. Pixelization is the mapping of each data point in some set to a pixel in a 2D image. Dimensional stacking is a layout method where N dimensions are projected onto the axes of an information display. We have combined and expanded upon both methods in an application named NeuroVis that supports interactive, visual data mining. Users can spontaneously perform ad hoc queries, cluster the results through dimension reordering, and execute analyses on selected pixels. While NeuroVis is not intrinsically restricted to any particular database, it is named after its original function: the examination of a vast neuroscience database. Images produced from its approaches have now appeared in the Journal of Neurophysiology and NeuroVis itself is being used for educational purposes in neuroscience classes at Emory University. In this paper we detail the theoretical foundations of NeuroVis, the interaction techniques it supports, an informal evaluation of how it has been used in neuroscience investigations, and a generalization of its utility and limitations in other domains.