One of the problems in visualizing multi-dimensional data is that the n-dimensions must be projected onto a two-dimensional display, often losing important information in the process. In addition, if an analyst intends rotating the n-dimensional cube, and projecting a different set of dimensions, then the projection must be made as quickly as possible, to keep the attention of the analyst through repeated projections.
This paper provides a fast projection technique for projecting n-dimensional cubes onto a two-dimensional display, called the nearest neighbor projection (NNP), that preserves neighborhood and clustering information. In addition, the authors provide a metric for evaluating a projection, or comparing two projections of the same dataset. Several sample and comparison projections are provided, in high quality graphics.
This is a high quality paper, and should be of interest to anyone who focuses on the mathematical and algorithmic aspects of data visualization.