It is well-observed that the whole world is full of data that are highly related and of diverse data object types such as people, organizations, and events. In many applications, it is intended to discover the hidden structures through such relationships involving different types of data objects in the world, in addition to "clusters" of the same type of data objects. On the other hand, relational data learning typically involves a large collection of data objects and thus algorithms for relational data learning are computation-intensive as well as data intensive. This calls for massively parallel solutions in order to make the algorithms scalable to large collections of data. This project addresses a three year integrated research and education program focusing on engaging in-depth research in developing novel parallel frameworks for a wide spectrum of state-of-the-art solutions to a series of fundamental problems in relational data learning. This research promotes the revolutionized understanding of relational data learning in the context of distributed computation environment. The project addresses fundamental problems in the literature of relational data learning as well as the expected breakthrough in the interdisciplinary and multidisciplinary research communities including parallel computation and scheduling, data mining and machine learning, and pattern analysis. The technologies generated from the research can be immediately deployed in important applications such as social network analysis, biological information discovery, financial and economic development analysis and prediction, natural disaster prediction, as well as military intelligence analysis.
Project url:
