Hendrik/xgboost
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Implement robust regularization in 'survival:aft' objective (#5473)

Browse Source 
* Robust regularization of AFT gradient and hessian

* Fix AFT doc; expose it to tutorial TOC

* Apply robust regularization to uncensored case too

* Revise unit test slightly

* Fix lint

* Update test_survival.py

* Use GradientPairPrecise

* Remove unused variables
This commit is contained in:
Philip Hyunsu Cho
2020-04-04 12:21:24 -07:00
committed by GitHub
parent 939973630d
commit 5fc5ec539d
9 changed files with 205 additions and 42 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
tests/python/test_survival.py
View File

@@ -85,6 +85,6 @@ def test_aft_survival_demo_data():
# AFT metric (negative log likelihood) improve monotonically
assert all(p >= q for p, q in zip(nloglik_rec[dist], nloglik_rec[dist][:1]))
# For this data, normal distribution works the best
assert nloglik_rec['normal'][-1] < 5.0
assert nloglik_rec['logistic'][-1] > 5.0
assert nloglik_rec['extreme'][-1] > 5.0
assert nloglik_rec['normal'][-1] < 4.9
assert nloglik_rec['logistic'][-1] > 4.9
assert nloglik_rec['extreme'][-1] > 4.9