Skip to content

Adding the possibility of doing a n-flavours fit

Created by: scarlehoff

Right now all fits have been done in a 8-flavours basis. In principle most of the code should be agnostic to the number of flavours that get out of the NN but there are two places that need to be changed:

  1. In the preprocessing file https://github.com/NNPDF/nnpdf/blob/master/n3fit/src/n3fit/layers/Preprocessing.py most of the code actually does not care about the number of flavours, only the call to the function, which can be equally well written as a loop. I think the only reason the loop is "unrolled" was to put the comments that make clear which preprocessing factor was what.

  2. I think the right rotation provided in pdfbases.py would be enough for the rest. The only other place where things can fail is at the leven of Rotation.py https://github.com/NNPDF/nnpdf/blob/master/n3fit/src/n3fit/layers/Rotations.py I am not sure how much it is relying on having 8 basis as the output of the NN at this point.

  3. The x operations in this file https://github.com/NNPDF/nnpdf/blob/master/n3fit/src/n3fit/layers/x_operations.py just need to be feed the right dimensionality. It might even work if the wrong number is given (i.e., not sure how useful is the output_dim keyword).