PNNL

The LEarning-Accelerated Domain Science (LEADS) institute has been selected as the newest Scientific Discovery through Advanced Computing (SciDAC) institute supported by the Department of Energy’s Advanced Scientific Computing Research program. Panos Stinis, leader of the Computational Mathematics group at Pacific Northwest National Laboratory, will direct the LEADS institute.

Stinis is joined on the LEADS leadership team with researchers from 14 different institutions across academia and the national laboratory complex, including LEADS Deputy Director Guannan Zhang from Oak Ridge National Laboratory.

“LEADS will have a transformative impact on research by directly integrating scientific machine learning into domain science challenges,” said Stinis. “Operator learning will enhance modeling for energy, environmental, and national security applications; graph-based artificial intelligence will efficiently analyze large-scale, complex datasets for the power grid and energy infrastructure; and digital-twin-assisted optimal control will enable real-time information extraction from user facility data.”

LEADS will complement the work of other SciDAC institutes, building upon the discretization methods and uncertainty quantification knowledge from FASTMath and the heterogeneous computing and data management expertise from RAPIDS to facilitate a paradigm shift within the field of scientific machine learning.

“LEADS will bridge the gap between scientific machine learning experts and domain scientists, enabling the development of state-of-the-art, highly customized, accurate, and efficient algorithms that leverage the vast domain knowledge within the Department of Energy complex,” said Stinis.

As with other SciDAC institutes, partnerships will play a key role in the success of LEADS. Domain scientists who wish to integrate scientific machine learning into their research should engage with the LEADS institute as part of DOE’s upcoming SciDAC partnership calls.