high performance computing on graphics processing units: hgpu.org

Significant advances have been made in time-varying data analysis and visualization, mainly in improving our ability to identify temporal trends and classify the underlying data. However, the ability to perform cost-effective data querying and indexing is often not incorporated, which posts a serious limitation as the size of timevarying data continue to grow. In this paper, we present a new approach that unifies data compacting, indexing and classification into a single framework. We achieve this by transforming the timeactivity curve representation of a time-varying data set into a hierarchical symbolic representation. We further build an indexable version of the data hierarchy, from which we create the iTree for visual representation of the time-varying data. A hyperbolic layout algorithm is employed to draw the iTree with a large number of nodes and provide focus+context visualization for interaction. We achieve effective querying, searching and tracking of time-varying data sets by enabling multiple coordinated views consisting of the iTree, the symbolic view and the spatial view.