Generalization of partitioned Runge-Kutta methods for adjoint systems

DOI10.1016/J.CAM.2020.113308MaRDI QIDQ2223878zbMATH OpenOpenAlexFDO

Authors T. Matsuda, Yuto Miyatake

Publication date 3 February 2021

Published in Journal of Computational and Applied Mathematics (Search for Journal in Brave)

Copyright license Creative Commons Attribution 4.0 International

Full work available at URL https://arxiv.org/abs/2003.09789

zbMATH Keywords

geometric integrationadjoint systemspartitioned Runge-Kutta method

Mathematics Subject Classification ID

Numerical optimization and variational techniques (65K10) Multistep, Runge-Kutta and extrapolation methods for ordinary differential equations (65L06) Numerical methods for Hamiltonian systems including symplectic integrators (65P10) Numerical solution of inverse problems involving ordinary differential equations (65L09)

Abstract: This study computes the gradient of a function of numerical solutions of ordinary differential equations (ODEs) with respect to the initial condition. The adjoint method computes the gradient approximately by solving the corresponding adjoint system numerically. In this context, Sanz-Serna [SIAM Rev., 58 (2016), pp. 3--33] showed that when the initial value problem is solved by a Runge--Kutta (RK) method, the gradient can be exactly computed by applying an appropriate RK method to the adjoint system. Focusing on the case where the initial value problem is solved by a partitioned RK (PRK) method, this paper presents a numerical method, which can be seen as a generalization of PRK methods, for the adjoint system that gives the exact gradient.

Recommendations

• Symplectic Runge-Kutta schemes for adjoint equations, automatic differentiation, optimal control, and more
• Evaluating gradients in optimal control: continuous adjoints versus automatic differentiation
• Runge-Kutta methods in optimal control and the transformed adjoint system
• On the Properties of Runge-Kutta Discrete Adjoints
• Automatic differentiation of explicit Runge-Kutta methods for optimal control

Cites work

• scientific article; zbMATH DE number 495924 (Why is no real title available?)
• Adjoint-based exact Hessian computation
• An introduction to the adjoint approach to design
• Deep learning
• Geometric Numerical Integration
• Partitioned Runge-Kutta Methods for Separable Hamiltonian Problems
• Spatially partitioned embedded Runge-Kutta methods
• Stable architectures for deep neural networks
• Symplectic Runge-Kutta schemes for adjoint equations, automatic differentiation, optimal control, and more
• The adjoint Newton algorithm for large-scale unconstrained optimization in meteorology applications
• The adjoint method in seismology

Cited in

(7)

• On the multisymplecticity of partitioned Runge–Kutta and splitting methods
• Adjoint-based inversion for stress and frictional parameters in earthquake modeling
• Partitioned Runge–Kutta–Chebyshev Methods for Diffusion-Advection-Reaction Problems
• Implicit peer triplets in gradient-based solution algorithms for ODE constrained optimal control
• On Multisymplecticity of Partitioned Runge–Kutta Methods
• Estimation of ordinary differential equation models with discretization error quantification
• scientific article; zbMATH DE number 124618 (Why is no real title available?)

Describes a project that uses

Uses Software

• torchdiffeq

This page was built for publication: Generalization of partitioned Runge-Kutta methods for adjoint systems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2223878)