The Data Discovery project is developing a collection of algorithms and software for the detection and characterization of information in massive data sets. This work is funded by the ASC Pre-and Post-Processing Environments project, and so is focused on simulation data sets in particular. The ideas and software have been successfully applied more broadly, however; for instance, to text and image analysis.
The department focuses on the "detection" half of the problem, and develops algorithms for robust, accurate pattern recognition in terascale data. The idea is that a user should be able to search their simulation data in a fashion similar to searching for text on the web. The "characterization" half of the problem, how to efficiently and richly describe the relevant regions found by the pattern recognition, is a project in the Viz and Scientific Computing department.
