Advisors : David Monniaux, Michaël Périn
Convex polyhedra appear in various areas in mathematics and computer science, including static analysis by abstract interpretation. The usual approaches for computing over convex polyhedra, as implemented in mainstream libraries, are based on computations in extended precision integer or rational numbers. This hampers efficiency ; we would often prefer to compute everything in floating-point. The goal of the thesis is to design algorithms over convex polyhedra in floating-point arithmetic.
This is a CIFRE PhD with The MathWorks.
This topic can also be started as a master’s internship.
(It is also possible to pursue such a PhD funded by the scholarships handed out competitively by the graduate school.)