High-Performance Statistical Computing: Building a Statistics Community within Supercomputing

Moderator: George Ostrouchov (Oak Ridge National Laboratory (ORNL); University of Tennessee, Knoxville)

Panelists: Marc G. Genton (King Abdullah University of Science and Technology (KAUST)), David Keyes (King Abdullah University of Science and Technology (KAUST), Columbia University), Norman Matloff (University of California, Davis), Douglas Nychka (Colorado School of Mines), Georgia Tourassi (Oak Ridge National Laboratory (ORNL)), Rio Yokota (Tokyo Institute of Technology)

Abstract: The current state of statistics and supercomputing is that the two communities are largely separate with little interaction. There is a confluence of developments that signals a timely need to foster more interaction and to build a statistics community within supercomputing. These developments include increased interest in explainable AI methods, calls to address potential bias in AI systems, interest in fast randomized algorithms for numerical computing and a need for greater accounting for uncertainty in predictions.

The panel will discuss some realities and barriers that may be responsible for the current situation and make recommendations on steps that can be taken to increase interaction and build a new community. This is the first such panel at SC and we plan to form a similar panel at CompStat or a similar statistics-centered conference.

Website: