Let models declare linear parameters to skip unnecessary Levenberg iterations

[div-dev123]

Hey @ltorres6 , fantastic work on the repository I have seen the repo last year too and its a massive improvement.

Since you mentioned thinking about enhancing the project as a whole ,I came across this issue:

According to my understanding , models have linear and non linear parameters. Non linear parameters are computed through iterations while linear are deterministic.

There are some instances like Paltak model (dce/models/paltak.py) which are iteratively fitted but its linear. the linear fitting is only used to calculate the starting point and then the iterator is run. This is uncessarily decreasing performace.
This is because the once fit is called it follows the same infrastructre of fit model , bound model and then levenberg fitter and we dont know if the model is linear or not.

Possible Solution:

The models could declare which parameters are linear and non linear and the levenberg fitter could have an "if" condition to check if a parameter is non linear and iterate over them .

Does this sound something worth looking over?

Regards
Divyansh Agrawal

Note: I used AI to understand the domain knowledge related to this and verify my findings