Developing novel algorithms and foundations for discrete structures, including AI, graph algorithms, and discrete optimization.
This group works to improve and expand the functionality of existing computational algorithms to make them compatible with the novel architectures of upcoming supercomputers, as well as to optimize them for fields of particular interest to DOE, including neuromorphic computing and quantum computing.
By developing advanced algorithms for high-performance computing systems, this group enables researchers to obtain results in seconds instead of days and thus accelerate the transition from simulations to practical experiments and real-world applications.
- Distributed ML theory and algorithms on HPC and emerging stack
- Summit scale HPC graph algorithms -- Breadth First Search, All Pairs Shortest Paths, etc.
- National security ML applications
- Safe AI
- Advances in Machine Learning to Improve Scientific Discovery at Exascale and Beyond (ASCEND)
- Scalable Non-Linear Dimensionality Reduction with Scientific Constraints
- Addressing GeoAI Low-shot Learning Privacy Challenges
- ExaLearn - ECP Co-Design Center
- Advanced Data SCiENce Toolkit for Non-Data Scientists (ASCENDS)
- Parallel Low Rank Approximation with Non-negativity Constraint (PLANC)
- Super DistLU
- Scientific Discovery through Advanced Computing (SciDAC)
- Adaptable IO System (ADIOS)
- Center for Online Data Analysis and Reduction (CODAR)
- URBAN-NET: Predicting Propagation Consequences Using Synergistically Interacting Infrastructure Networks
- CODAR: ECP Center for Online Data Analysis and Reduction