search

NNPDF / nnpdf

An open-source machine learning framework for global analyses of parton distributions.
https://docs.nnpdf.science/
GNU General Public License v3.0
29 stars 5 forks source link

Kinematic coverage for polarized datasets #2104

Open scarlehoff opened 3 weeks ago

scarlehoff commented 3 weeks ago

While in principle a process type was added, the implementation of the (unpolarized) DIS datasets are slightly different from the polarized ones. Mainly because of the missing kinematic variables (DIS still use k1/k2/k3).

https://github.com/NNPDF/nnpdf/blob/6691f56f975f27732a9102b8ef58b01b04f2b810/validphys2/src/validphys/process_options.py#L124

It might be enough to modify this function to

    x = kin_info.get_one_of("k1", _Vars.x)
    q = kin_info.get_one_of("k2", _Vars.Q)
    return x, q * q

but it requires some testing to make sure that the results are as expected.

Radonirinaunimi commented 3 weeks ago

I have already looked into this and fixed it before to generate kinematic plots for a talk (slack-slides). But I couldn't find anymore nor remember what has done back then.

But your suggestion (with a very tiny tweak) works in https://github.com/NNPDF/nnpdf/pull/2109 and gives the same result as back then (report).