Does declaring jacobian and redefining ODEProblem with jacobian accelerate `solve`?

General Usage
,
1

https://diffeq.sciml.ai/dev/features/performance_overloads/#ode_explicit_jac

following from this example,…
Does explicitly declaring jacobian for an ODEproblem like this accelerate the solution using solve()

2

It can, if you write a better version that autodiff would build.

1 Like
3

@ChrisRackauckas

Like in case of symbolic generated Jacobian using MTK? that is better than AutoDiff, right?

In case of Reactionsystem converted to ODESystem, we do not have generic function f used for creating odeproblem with Jacobian right?? In such case, how do we assign this Jacobian to converted ODESystem?

4

Not necessarily. Depends on whether you can get simplifications or exploit sparsity.

You do.

You just generate it with jac=true at problem construction time.

5

You just generate it with jac=true at problem construction time.

Like a kwarg as follows ?

oprob = ODEProblem(odesys, x₀map, tspan, pmap;jac=true);
6

Yup, just as the docstring describes.

https://mtk.sciml.ai/dev/systems/ODESystem/#SciMLBase.ODEProblem