Description

Overflow Runtime Warning for Numpy Logistic Function in CalibratedClassifierCV with SVM

Steps/Code to Reproduce

Training a soft-margin SVM inside a CalibrationClassifierCV using sigmoid (Platt) scaling. Below is the definition of the classifier:

g_svm = svm.SVC(gamma='auto') g = CalibratedClassifierCV(g_svm, method='sigmoid', cv=10)

I am not able to share the dataset publicly. I could try to create a synthetic one if it is useful/needed. The warning is intermittent.

Expected Results

No runtime warning is reported.

Actual Results

/sklearn/calibration.py:520: RuntimeWarning: overflow encountered in exp
return 1. / (1. + np.exp(self.a_ * T + self.b_))

Versions

System:
python: 3.7.1 (default, Dec 12 2018, 01:02:44) [Clang 10.0.0 (clang-1000.11.45.5)]
executable:
machine: Darwin-18.2.0-x86_64-i386-64bit

BLAS:
macros: NO_ATLAS_INFO=3, HAVE_CBLAS=None
lib_dirs:
cblas_libs: cblas

Python deps:
pip: 18.1
setuptools: 39.0.1
sklearn: 0.20.1
numpy: 1.15.4
scipy: 1.1.0
Cython: None
pandas: 0.23.4