Numerical Linear Algebra Group

The Gilles Kahn prize was awarded to Theo Mary for his PhD thesis Block Low-Rank multifrontal solvers: complexity, performance, and scalability (pdf here). The thesis was prepared at the University of Toulouse, France and co-supervised by Patrick Amestoy and Alfredo Buttari. Theo is currently a Research Associate in the Numerical Linear Algebra group.

The Gilles Kahn prize is awarded each year by the SIF, the French Society of Computer Science, for an excellent PhD thesis in the field of computer science. 

In the thesis, Theo investigated the use of low-rank approximations inside direct, factorization-based methods to reduce the computational cost of the solution of large, sparse systems of linear equations. His work included theoretically computing the asymptotic complexity reduction achieved by these methods and implementing them on parallel computers to translate this theoretical reduction into significant performance gains on both shared- and distributed-memory architectures. The potential and efficiency of these methods was assessed on several industrial applications, notably ones arising in geosciences and structural mechanics.

The official award ceremony will take place at the SIF congress (February  6-7, 2019, in Bordeaux). Theo has also been invited to present his work to the entire scientific community at the SIF congress.