SUPR
Creating new ML numerical ODE/ODE solvers
Dnr:

NAISS 2023/22-1305

Type:

NAISS Small Compute

Principal Investigator:

Alexandros Sopasakis

Affiliation:

Lunds universitet

Start Date:

2024-01-01

End Date:

2025-01-01

Primary Classification:

10105: Computational Mathematics

Webpage:

Allocation

Abstract

Different types of machine learning Autoencoders are to be explored in order to create new methods for solving ordinary and partial differential equations. In particular we explore solvers for time dependent differential equations where there is significant luck in solutions of ML type. Also and as a more challenging step stochasticity will be explored in such differential equations and solvers for those will also be considered. Solvers to be developed will be learning on data but also from a combination of data and physics informed dynamics.