CA angles fit #9
Description
In the new carbon-alpha support we just merged, angles are by default fit with a harmonic function.
In the tiny demo dataset this results in a failure to fit the angles (failed to fit potential estimate for harmonic)
In the big charmm trpcage dataset, the fit works (and also produces a good model according to @sayeg84) but taking a look at the plots, it's clear that the harmonic fit is not ideal.
On the other hand, a bimodal fit is often less stable and requires more hand-tweaking.
It would be nice to have a new prior class for bimodal angles (maybe in mlcg directly) and then have the corresponding fitting function here.
that way we would have the option of using it, even if the default should remain the harmonic as it works for most of the systems we are investigating and throws less errors.