Skip to content

Change formatting of Training length hist and T/V scatter plots

Emanuele Roberto Nocera requested to merge equal-box-tl-plot into master

Created by: wilsonmr

Before: https://vp.nnpdf.science/qa-7Pw-lSzGK8kVK2UOU_Q==/

  • TL plot was "density" hist but I personally don't think we want a PDF here because it includes bin width in integral and instead I think a PMF of proprtion of replicas in each bar is more useful scale.
  • T/V scatter has different x and y axis so "skewness" comparisons are hard to make

After: https://vp.nnpdf.science/4rY-pmBMRX-n3SSL7HgP_g==/

  • TL plot changed as described above, now can instantly see e.g that ~10% of replicas went to ~max training length
  • T/V scatter now has equal axis, I set height and width to default width so it's a bit taller and I add a line of y=x for reference

Is there a historical reason for doing this? Otherwise does anyone hate it? I think it's easier to understand.

Merge request reports