Jacobian definition about parameterizedfunctions.jl HOT 6 CLOSED

Nope, we now use ODEFunction.

from parameterizedfunctions.jl.

Ok, so if I was using my own Parameterized function type, I need to adopt the ODEFunction interface, and instead subtype AbstractODEFunction?

from parameterizedfunctions.jl.

from parameterizedfunctions.jl.

The usecase is solving a Quantum Liouvillian, where the operators are sparse and complex, and I use the 5-arg mul! to have everything in place because I also need to evaluate time-dependent user functions as coefficients for some of the operators (see here).

As far as I can tell, none of the pre-existing solutions work for this, which is why I made my own type.

from parameterizedfunctions.jl.

I'd like to talk more about getting this to work if you'd have the time,

w.r.t. this issue, it seems like ParameterizedFunctions are on the way out and that the (p::LotkaVolterra)(::Type{Val{:jac}},t,u,J) syntax is not supported anymore. Is that correct?

In case it matters, it would be nice to still support it, for those who like me are using functor types. With functor types, it is impossible to define a function in the jac field, because the function would itself depends on the instance of the functor...

from parameterizedfunctions.jl.

w.r.t. this issue, it seems like ParameterizedFunctions are on the way out and that the (p::LotkaVolterra)(::Type{Val{:jac}},t,u,J) syntax is not supported anymore. Is that correct?

Yes, that's been gone for awhile.

In case it matters, it would be nice to still support it, for those who like me are using functor types. With functor types, it is impossible to define a function in the jac field, because the function would itself depends on the instance of the functor...

You can define an ODEFunction and add your own jac

The usecase is solving a Quantum Liouvillian, where the operators are sparse and complex, and I use the 5-arg mul! to have everything in place because I also need to evaluate time-dependent user functions as coefficients for some of the operators (see here).

That's fine with MTK.

from parameterizedfunctions.jl.