L'objet de ce project est l'analyse numérique et la mise en oeuvre efficace d'algorithmes qui utilisent des techniques de tenseurs de faible rang, appliquées à trois types de problèmes :
Numerical methods using low-rank tensor techniques have seen great success in, for example, theoretical physics for the simulation of spin systems, in scientific computing for high-dimensional PDEs, and in signal processing for independent component analysis. The aim of this proposal is to (A) propose numerical integrators for integration of time-reversible tensor ODEs with rank truncation, (B) parallelize low-rank tensor algorithms for optimization and time integration, and (C) apply tensor techniques for regression and decision in machine learning.
Last updated:04.06.2022
