TST: test_predict_proba_binary fails for some random seed #12762
Description
Description
test_predict_proba_binary fails with some random seed. I encountered this during PR #10058 , modifying by mistake the random state in tests
Steps/Code to Reproduce
from sklearn.neural_network.tests.test_mlp import test_predict_proba_binary
np.random.seed(2)
test_predict_proba_binary()Expected Results
No error is thrown
Actual Results
---------------------------------------------------------------------------
AssertionError Traceback (most recent call last)
<ipython-input-1-54c5d3efee13> in <module>
1 from sklearn.neural_network.tests.test_mlp import test_predict_proba_binary
2 np.random.seed(2)
----> 3 test_predict_proba_binary()
~/anaconda3/envs/env_test_mlp/lib/python3.7/site-packages/sklearn/neural_network/tests/test_mlp.py in test_predict_proba_binary()
448 assert_array_equal(y_log_proba, np.log(y_proba))
449
--> 450 assert_equal(roc_auc_score(y, y_proba[:, 1]), 1.0)
451
452
~/anaconda3/envs/env_test_mlp/lib/python3.7/unittest/case.py in assertEqual(self, first, second, msg)
837 """
838 assertion_func = self._getAssertEqualityFunc(first, second)
--> 839 assertion_func(first, second, msg=msg)
840
841 def assertNotEqual(self, first, second, msg=None):
~/anaconda3/envs/env_test_mlp/lib/python3.7/unittest/case.py in _baseAssertEqual(self, first, second, msg)
830 standardMsg = '%s != %s' % _common_shorten_repr(first, second)
831 msg = self._formatMessage(msg, standardMsg)
--> 832 raise self.failureException(msg)
833
834 def assertEqual(self, first, second, msg=None):
AssertionError: 0.5 != 1.0
Versions
System:
python: 3.7.1 (default, Oct 23 2018, 19:19:42) [GCC 7.3.0]
executable: /home/will/anaconda3/envs/env_test_mlp/bin/python
machine: Linux-4.4.0-139-generic-x86_64-with-debian-stretch-sid
BLAS:
macros: SCIPY_MKL_H=None, HAVE_CBLAS=None
lib_dirs: /home/will/anaconda3/envs/env_test_mlp/lib
cblas_libs: mkl_rt, pthread
Python deps:
pip: 18.1
setuptools: 40.6.2
sklearn: 0.21.dev0
numpy: 1.15.4
scipy: 1.1.0
Cython: None
pandas: None