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

Fix #3397: early_stop callback does not maximize metric of form NDCG@n- (#3685)

Browse Source 
* Fix #3397: early_stop callback does not maximize metric of form NDCG@n-

Early stopping callback makes splits with '-' letter, which interferes
with metrics of form NDCG@n-. As a result, XGBoost tries to minimize
NDCG@n-, where it should be maximized instead.

Fix. Specify maxsplit=1.

* Python 2.x compatibility fix
This commit is contained in:
Philip Hyunsu Cho 2018-09-08 19:46:25 -07:00 committed by GitHub
parent f606cb8ef4
commit 3564b68b98
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 5 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
python-package/xgboost/callback.py
View File

@ -191,19 +191,18 @@ def early_stop(stopping_rounds, maximize=False, verbose=True):
maximize_metrics = ('auc', 'map', 'ndcg')
maximize_at_n_metrics = ('auc@', 'map@', 'ndcg@')
maximize_score = maximize
metric = env.evaluation_result_list[-1][0]
metric_label = env.evaluation_result_list[-1][0]
metric = metric_label.split('-', 1)[-1]
if any(env.evaluation_result_list[-1][0].split('-')[-1].startswith(x)
for x in maximize_at_n_metrics):
if any(metric.startswith(x) for x in maximize_at_n_metrics):
maximize_score = True
if any(env.evaluation_result_list[-1][0].split('-')[-1].split(":")[0] == x
for x in maximize_metrics):
if any(metric.split(":")[0] == x for x in maximize_metrics):
maximize_score = True
if verbose and env.rank == 0:
msg = "Will train until {} hasn't improved in {} rounds.\n"
rabit.tracker_print(msg.format(metric, stopping_rounds))
rabit.tracker_print(msg.format(metric_label, stopping_rounds))
state['maximize_score'] = maximize_score
state['best_iteration'] = 0