search

ogrisel / pygbm

Experimental Gradient Boosting Machines in Python with numba.
MIT License
183 stars 32 forks source link

split_info comparison similar to LGBM #17

Closed NicolasHug closed 6 years ago

NicolasHug commented 6 years ago

In LGBM when the gains are the same, the best split info is the one that has the highest lowest feature (link)

This currently leads to small differences in the splits.

EDIT: closing... This fixes one of the discrepancies I found, but probably not for the good reasons

codecov-io commented 6 years ago

Codecov Report

Merging #17 into master will increase coverage by 0.02%. The diff coverage is 100%.

Impacted file tree graph

@@            Coverage Diff             @@
##           master      #17      +/-   ##
==========================================
+ Coverage   93.13%   93.16%   +0.02%     
==========================================
  Files           8        8              
  Lines         714      717       +3     
==========================================
+ Hits          665      668       +3     
  Misses         49       49
Impacted Files Coverage Δ
pygbm/grower.py 89.94% <100%> (+0.18%) :arrow_up:
pygbm/splitting.py 100% <100%> (ø) :arrow_up:

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 5e30c96...fd38d17. Read the comment docs.