Harmonize Constant/FixedParameter/ConstantParameter

[FFroehlich]

happy to discuss what they should be called.

condition
feature
constant

[dweindl]

I am also in favor of finding more appropriate names, as it is often confusing. No good solution, just some thoughts:

What is different for those parameters and others? We don't compute sensitivities with respect to those. So they are generally not estimated, but known. Is "input parameter" appropriate? Somehow. But it may also include known output parameters, making it at least sound somewhat contradictory.

"Condition parameters": Often makes sense, as they are used to account for different conditions a model should represent. On the other hand, we may want to estimate condition parameters as well, and therefore they would be in the non-condition group.

"feature": Not very descriptive. Could be anything?

"constant": All model parameters are constant over time. These parameters are not constant over conditions. They are constant for one condition over the course of the optimization. Makes sense, but not sure if it is the first association, as AMICI itself does not care greatly about optimization.

Other "types" of parameters we may want to distinguish, orthogonal to the sensitivity thing, but maybe good to consider to avoid ambiguities: parameter influencing the dynamics of the system vs output-only parameters.

[dweindl]

Any other opinions here?

[FFroehlich]

The reason why I was thinking about feature is that those fixed parameters, as you say, really could be anything related to the experiment. If one would train a machine learning model, these quantities would be called features.