The way we handle parameters

[sumiya11]

I propose we discuss the way we treat parameters in ODEs here, so that it doesn't get lost 🐡

The algorithm does not distinguish parameters from state variables. Though a polynomial ring variable can be marked as a parameter in the ODE, we implicitly assign a zero derivative to it.

The result of the algorithm is again an ODE system, which, at the moment, has no parameters. The proposition is do declare some of the state variables of the new ODE as parameters according to some strategy. Two already discussed strategies are: everything constant is a parameter; a combination of parameters is a parameter.

From the users perspective , the difference is that zero equations for parameters are not printed, as in:

@ODEsystem(
     x'(t) = a*x + 1
 )
[ Info: Summary of the model:
[ Info: State variables: x
[ Info: Parameters: a
x'(t) = x(t)*a(t) + 1

Additionaly, there is a consensus that we want a seamless transition between our ODE and MTK ODESystem, if possible. Currently, conversion works via ODEtoMTK and MTKtoODE, it transforms state variables into state variables and parameters into parameters.

cc @pogudingleb

[pogudingleb]

I will try to work out a case with MTK to have a more informed opinion.

Two thought so far:

• a parameter should not be printed with (t), this looks more like an input and would be quite confusing.
• in principle, we could make the parameter-choosing strategy to be a parameter of the algorithm.

[sumiya11]

a parameter should not be printed with (t)

Good point

[sumiya11]

@pogudingleb any objections to closing this ?

[pogudingleb]

Nope