From 79a860b0a7a6d420554195545b28f969ba20e182 Mon Sep 17 00:00:00 2001
From: Olivier Grisel <olivier.grisel@ensta.org>
Date: Thu, 21 Oct 2010 21:04:44 +0200
Subject: [PATCH] FIX: make grid_search output deterministic even in case of
tie on the scores
---
scikits/learn/grid_search.py | 15 ++++++++++++++-
1 file changed, 14 insertions(+), 1 deletion(-)
diff --git a/scikits/learn/grid_search.py b/scikits/learn/grid_search.py
index 1d8ca1378f..22e27c0d97 100644
--- a/scikits/learn/grid_search.py
+++ b/scikits/learn/grid_search.py
@@ -210,7 +210,20 @@ class GridSearchCV(BaseEstimator):
for clf_params in grid)
# Out is a list of pairs: score, estimator
- self.best_score, best_estimator = max(out) # get maximum score
+
+ # Note: we do not use max(out) to make ties deterministic even if
+ # comparison on estimator instances is not deterministic
+ best_score = None
+ for score, estimator in out:
+ if best_score is None:
+ best_score = score
+ best_estimator = estimator
+ else:
+ if score >= best_score:
+ best_score = score
+ best_estimator = estimator
+
+ self.best_score = best_score
if refit:
# fit the best estimator using the entire dataset
--
GitLab