Objective functions #7
Description
Problem
In the original pipeline, we defined the config like this in the runner cell:
cfg = RefinementConfig(
rpoly=1, hiQ_frac=.85, ridge=1e-3, Qmax=50.0,
window="none", rmin=0.0, rmax=20.0, dr=0.01, #lorch/kaiser/none
n_outer=24, plot_every=2, frame_dir="./refine_frames"
)
Our data only goes up to approx. 30 A^-1. When I change the Qmax in the config file, the pipeleline behaves a bit strangely. If you look at the frames, it does a good job up to frame 18, but then it just decides to completely flatten out S(Q). This only happens when I set Qmax to 30, instead of 50. The Qmax defined in the config does not influence the polynomial degree, which takes Qmax from the maximum value in the Q array. It only affects how the Lorch and Kaiser functions are defined.
Proposed solution
Was the a specific logic why we had set Qmax to 50? If so, we need to rationalize it, or make the code work well with the "real" Qmax. We probably need to change the objective functions. Modifying the "lamda_smooth" threshold to a lower number in the config file helps, for example.