Randomized Algorithms for Scientific Computing (RASC)


DOE RASC Workshop Announcement

Dear colleagues,

We are excited to announce a virtual workshop sponsored by the US Department of Energy (DOE) Advanced Scientific Computing Research (ASCR) Applied Mathematics program:

RASC: Randomized Algorithms for Scientific Computing
(an extended virtual workshop in two parts)
Part 1: December 2-3, 2020 @ 11am-3pm Eastern Time
Part 2: January 6-7, 2021 @ 11am-3pm Eastern Time

New high-performance algorithms, models, and simulations are crucial for US leadership in science (The Future of Computing, Industries of the Future, Future Advanced Computing Ecosystems, AI for Science). The purpose of this workshop is to explore what foundational long-term research and plans are needed over the next decade to transform DOE’s scientific computing capabilities, employing randomized algorithms (e.g., random sampling, randomized projections, stochastic gradient descent, randomized numerical linear algebra) to achieve marked gains in performance for scientific computing grand challenges. Potential topics of interest to address through the use of randomness include the following:

High computational complexity and the development of efficient algorithms in large-scale simulations High data dimensionality and finding sparse representations for data from user facilities Better algorithm scalability for low-power, high-performance edge computing Reduced ill-conditioning and sensitivity for inverse problems Improved algorithm reliability and robustness to noise, including in nondeterministic post-Moore computing environments

Part 1 (Dec 2-3, 2020): Randomized Algorithms and Applications Bootcamp - Focused on providing a common understanding on the current status and the potential of randomized algorithms and applications within DOE that they may benefit. Features tutorial presentations and discussions with leading experts.

Part 2 (Jan 6-7, 2021): Brainstorming and Writing Sessions - Focused on attendee brainstorming and idea workshopping to develop a report on long-term research needs to realize the potential of randomized algorithms for scientific computing in the context of DOE applications.

The workshop is open to all, but registration is required. Registration for Part 1 is now open at https://www.orau.gov/rasc2020. Registration for Part 2 will open in December and requires registration for and participation in Part 1. Part 2 also requires the submission of a 200-word idea on long-term research needs to inform the discussion and the final report.

Aydin Buluc (LBNL), Tammy Kolda (SNL), Stefan Wild (ANL)
RASC Workshop Co-Chairs