Closed Moyoxkit closed 3 years ago
You'll need to format with black before merging.
Some other minor comments for you
I've updated the penalty functions I added to now do the add in squares technique we discussed last Friday, still with an L2 (Gaussian-like) norm. Also implemented your comments.
Excellent thanks!
This push adds four new Gaussian penalty functions:
Based around observational errors, a constant percent error, A mix of the two, where the largest is always taken and a variant of the mixed function with an additional weight factor that could help combine different observables.
As these models don't have a natural cutoff, I added a free parameter thats sets the amount of sigmas where the penalty will become 1.
Ive tested them all and they work as intended, I would be happy to use shorter names but they will be less descriptive. All these are intended for use for MCMC methods, where log(1-Penalty) can be used as a Gaussian log likelihood.
Also adds markers at the data points when using plot penalty