Scalable Algorithms for Graph Partitioning, Data Mining, and Cheminformatics
This research covers a wide-range of areas including parallel graph partitioning and clustering, runtime systems for automated out-of-core execution of MPI programs, parallel sparse tensor factorization, chemical informatics, and data mining. The research on parallel graph partitioning and clustering focuses on the multilevel paradigm and is designed to develop algorithms that can operate efficiently on systems containing from just a few many-core processors to systems with millions of processors. The research on out-of-core runtime systems focuses on developing a custom runtime for the standard MPI specification that will allow it to operate efficiently on problems whose aggregate amount of memory requirements exceeds the aggregate amount of DRAM available in the system. The research on parallel sparse tensor factorization is designed to develop computation and memory efficient algorithms for multi-mode tensors. The research in chemical informatics is designed to develop graph-theory and machine-learning based approaches for identifying the compounds that can modulate the function of specific proteins. Finally, the research on data mining is designed to develop fundamental algorithms for key problems that relate to analyzing dynamic networks, multivariate time series, identifying similar objects in sparse high dimensional datasets, multi-task and transfer learning and develop solutions in the areas of recommender systems, learning analytics, and educational data mining.
Return to this PI's main page.