Discovering Concepts in Structural Data

The explosive growth of databases in scientiic, industrial, and commercial elds has not been accompanied by a similar growth in our ability to analyze and digest this data. The increasing amount and complexity of data creates an urgent need for automatic database analysis tools. This trend is evident in molecular biology data which continues to grow in both size and complexity. This research outlines a general approach to automatically discover repetitive and functional concepts in large structural databases. The Subdue system discovers substructures that compress the database and represent structural concepts in the data. By replacing previously-discovered substructures in the data, multiple passes of Subdue produce a hierarchical description of the structural regularities in the data. To increase the exibility of the system, we describe methods of incorporating domain-dependent information into the discovery process. Because discovery systems such as Subdue are very computationally expensive, we also explore ways of parallelizing the system to improve scalability.