Operator networks are used as deep learning tools for approximating solutions of partial differential equations (PDEs). Robust methods for solving PDEs have already been developed, such as the finite element method (FEM), but they have a high computational cost. The architecture of a variationally mimetic operator network (VarMiON) mimics the form of the numerical solution obtained from an approximate variational or weak formulation of a PDE making it possible to obtain an efficient method that can solve multiple instances of a problem with different input functions. The literature on operator networks has already considered the steady-state heat equation. This thesis explores the problem of solving the time-dependent heat equation with a VarMiON.

Operator networks are used as deep learning tools for approximating solutions of partial differential equations (PDEs). Robust methods for solving PDEs have already been developed, such as the finite element method (FEM), but they have a high computational cost. The architecture of a variationally mimetic operator network (VarMiON) mimics the form of the numerical solution obtained from an approximate variational or weak formulation of a PDE making it possible to obtain an efficient method that can solve multiple instances of a problem with different input functions. The literature on operator networks has already considered the steady-state heat equation. This thesis explores the problem of solving the time-dependent heat equation with a VarMiON.

Variationally mimetic operator neural networks for the time-dependent heat equation

DELL'ORTO, MARCO
2023/2024

Abstract

Operator networks are used as deep learning tools for approximating solutions of partial differential equations (PDEs). Robust methods for solving PDEs have already been developed, such as the finite element method (FEM), but they have a high computational cost. The architecture of a variationally mimetic operator network (VarMiON) mimics the form of the numerical solution obtained from an approximate variational or weak formulation of a PDE making it possible to obtain an efficient method that can solve multiple instances of a problem with different input functions. The literature on operator networks has already considered the steady-state heat equation. This thesis explores the problem of solving the time-dependent heat equation with a VarMiON.
2023
Variationally mimetic operator neural networks for the time-dependent heat equation
Operator networks are used as deep learning tools for approximating solutions of partial differential equations (PDEs). Robust methods for solving PDEs have already been developed, such as the finite element method (FEM), but they have a high computational cost. The architecture of a variationally mimetic operator network (VarMiON) mimics the form of the numerical solution obtained from an approximate variational or weak formulation of a PDE making it possible to obtain an efficient method that can solve multiple instances of a problem with different input functions. The literature on operator networks has already considered the steady-state heat equation. This thesis explores the problem of solving the time-dependent heat equation with a VarMiON.
neural networks
heat equation
time-dependent
File in questo prodotto:
File Dimensione Formato  
Variationally_mimetic_operator_neural_networks_for_the_time_dependent_heat_equation_pdfa.pdf

accesso aperto

Dimensione 2.3 MB
Formato Adobe PDF
2.3 MB Adobe PDF Visualizza/Apri

The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.12608/64777