Discovery Visualization Using Fast Clustering

To attack the problem of handling increasingly vast stores of information, we discuss a new approach to data exploration that requires the close coupling of man and machine. We call this approach discovery visualization to emphasize the importance of visual display and interaction. This approach aims to discover new relations, new features, and new knowledge. A key element in discovery visualization lies in heightening the machine's awareness of users so they have, for example, focus-based manipulation, based on where and how closely they look at the displayed scene, in addition to direct manipulation. This process makes no sense unless the machine can respond immediately. Further, we promote the concept of continuous interaction with constant feedback between man and machine, and constant unfolding of the data. Finally, automated response must combine with user selection to achieve and sustain animated action, even in data sets of great or varying complexity.