The error control in DifferentialEquations using the default internal norm is as follow: Compute tolerance on i-th component: tolerance_i = absoluteTolerance_i + relativeTolerance_i * max( \lvert previousState_i \rvert, \lvert newState_i \rvert) Then compute the “normalised error vector”: erro…

[image] Bernard_GODARD: With this default choice, it means that if only one component has significant error that error is allowed to grow to sqrt(N) (N vector dimension) times the tolerance on that component before the step is reduced. It’s been the default choice since the 70’s, with Gear, S…

Default norm used in ODE error control

Specific Domains Modelling & Simulations

ranocha November 6, 2021, 6:13pm 4

You can pass keyword argument internalnorm, see https://diffeq.sciml.ai/stable/basics/common_solver_opts/#advanced_adaptive_stepsize_control

Topic		Replies	Views
Stepsize control in ODEProblem Numerics diffeq	11	3705	March 19, 2018
L2 vs. L2 error in Differential Equations Benchmarks New to Julia	3	870	October 18, 2019
Weird notation General Usage	2	476	February 7, 2017
DifferentialEquations.jl default options for a solver Numerics diffeq	2	2397	June 13, 2019
Norm, normalize Internals & Design	6	9719	February 1, 2017

Default norm used in ODE error control

Related topics