Jarník’s Lecture by Dr. Erin Carson

Erin Claire Carson is affiliated with the Department of Numerical Mathematics. She studied computer science at the University of Virginia and obtained her Ph.D. from the University of California, Berkeley. She spent three years at New York University. Since 2018, she has been an assistant professor at Charles University. Her research lies at the intersection of numerical linear algebra, high-performance computing, and parallel algorithms. In 2023, she received the prestigious ERC Starting Grant, under which she is developing new algorithms for cutting-edge supercomputers as part of a five-year project.

Abstract of the Lecture:

On supercomputers that exist today, achieving even close to the peak performance is incredibly difficult if not impossible for many applications. Techniques designed to improve the performance of matrix computations – making computations less expensive by reorganizing an algorithm, making intentional approximations, and using lower precision – all introduce what we can generally call “inexactness”. The questions to ask are then:

1. With all these various sources of inexactness involved, does a given algorithm still get close enough to the right answer?
2. Given a user constraint on required accuracy, how can we best exploit and balance different types of inexactness to improve performance?

Studying the combination of different sources of inexactness can thus reveal not only limitations, but also new opportunities for developing algorithms for matrix computations that are both fast and provably accurate. We present a few recent examples of this approach, in which mixed precision computation is combined with other sources of inexactness.

OPMK, photo by Tomáš Rubín