Sparse Automatic Differentiation
Alexis Montoison (Argonne National Lab.)
Mercredi 8 janvier 2025 à 11:00, salle Coriolis (bât. Galois)
Abstract. While traditional AD is well-integrated into high-level programming languages, automatic sparse differentiation (ASD) remains underutilized due to its origins in low-level programming research and graph theory. This presentation demystifies ASD by explaining its key components, such as sparsity pattern detection, matrix coloring, and their roles in the computation of both sparse Jacobians and Hessians.
Supported by PDE-AI projet of PEPR IA
