Closed redeboer closed 3 years ago
Sounds good to me.
A next step would of course be to build some more intelligence in the model building. But there we have to be careful not to tailor too much to specific analysis cases.
This is a very good point, not to tailor things too much. At least this requires very good knowledge which couplings are guaranteed to be correct. For example I'm not sure if generally the coupling of the widths for a particle and antiparticle is correct (C violations).
I think it's possible to do this since ComPWA/expertsystem#454, but it has to be illustrated in some notebook.
The
FitParameters
class is currently rather (too) simple: it only provides functionality to search for parameters and to add them. This is a problem if one wants to couple fit parameters.A simple solution would be to make have the
FitParameters
derive from aMutableMapping
and implement the required methods. One can then do something like:which is, I think (@spflueger?), sufficient to couple those parameters in
tensorwaves
.A next step would of course be to build some more intelligence in the model building. But there we have to be careful not to tailor too much to specific analysis cases.