NegLogParameterPriors does not evaluate prior density at fixed params

[arrjon]

Based on #1412

Making the NegLogParameterPriors aware of fixed parameters. The density is not evaluted for those. Moreover, the petab importet is updated to indicate the fixed parameters while creating the prior.

Question: should this argument stay optional? If priors are not given for fixed parameters, no problem would occure. The problem only occurs when priors are given and parameters fixed.

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 50.00000% with 3 lines in your changes missing coverage. Please review.

Project coverage is 83.90%. Comparing base (2250dbc) to head (b178863).

Files	Patch %	Lines
.../objective/roadrunner/petab_importer_roadrunner.py	50.00%	2 Missing ⚠️
pypesto/petab/importer.py	50.00%	1 Missing ⚠️

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #1413      +/-   ##
===========================================
- Coverage    83.92%   83.90%   -0.03%     
===========================================
  Files          160      160              
  Lines        13061    13067       +6     
===========================================
+ Hits         10962    10964       +2     
- Misses        2099     2103       +4     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[dilpath]

Thanks, I think it looks fine, but since we have prior_list, we can overwrite immediately, instead of recomputing it every time it's accessed.

[dilpath]

Thanks! Sorry for the mask suggestion, I wasn't familiar with the code

[PaulJonasJost]

Nice, a test for this would be good. Also not entirely sure whether it is relevant, but the problem has a function fix_parameters, for this it should be accounted as well? 🤔

[arrjon]

Also not entirely sure whether it is relevant, but the problem has a function fix_parameters, for this it should be accounted as well? 🤔

That is relevant, but the function calls normalize, which in turn calls update_from_problem, so no need to add in fix_parameters.

[arrjon]

After discussing this issue in #1412, we will not make the prior aware of fixed parameters. Instead, users should specifiy priors correctly in the PEtab problem. In this way, we will allow maximum flexibility.