af4a2ca795
DOC: Fix typo in CalibratedClassifierCV
2015-10-06 15:41:15 +02:00
6966881328
MAINT use inspect.signature for introspection
2015-09-23 16:30:53 +02:00
882c346abd
DOC Make cv documentation consistent across our codebase
2015-09-10 09:15:13 +05:30
87aaabc06a
DOC backticks in attribute docstrings unnecessaru since #3489
2015-06-12 05:13:02 +10:00
62325cbb58
MAINT merge _check_cv into check_cv as indices argument is removed in 0.17
2015-06-08 16:37:03 +05:30
0650d5502e
DOC adding backlinks to docstrings
2015-06-03 00:24:04 -04:00
2b666873d2
FIX make CalibratedClassifierCV deterministic by default.
2015-04-06 17:57:59 -04:00
05100fc2a1
FIX : allow NaN in input of calibration if estimator handles it
2015-03-18 16:44:21 +01:00
c1fa16f5be
ENH no need for tie breaking jitter in calibration
...
The isotonic regression routine now implements deterministic
tie-breaking by default.
2015-03-04 12:04:35 -05:00
70d49de695
ENH ensure that a warning is raised when sample_weight is not supported
...
Also: make it possible to fix the random_state used to break the jitter
for isotonic calibration.
2015-02-20 19:37:12 +01:00
eee0e67fa6
MAINT mark _sigmoid_calibration private
2015-02-20 19:37:12 +01:00
7376309e81
TST improve coverage of calibration.py
2015-02-20 19:37:11 +01:00
6b0f75ccc5
COSMIT : pep8 + 2 spaces
2015-02-20 19:37:11 +01:00
ecfc93ddf6
ENH Add probability calibration based on isotonic regr. and Platt's sigmoid fit + calibration-curve
...
CalibratedClassifierCV allows to calibrate the predicted probabilities of base classifiers based on a cross-validation scheme and either Platt's sigmoid fit or isotonic regression. This can be used to compensate for an under-confident or over-confident classifier. It allows also to turn the decision scores of a non-probabilistic classifier into valid probabilities.
The function calibration_curve allows to evaluate how well calibrated the probabilties returned by a classifier are. Ideally, the curve should be close to diagonal.
2015-02-20 19:37:10 +01:00
Allen Riddell