7e29af034b
MAINT remove deprecated stuff that will no longer be supported in 0.18
2015-10-19 10:44:42 +02:00
664d78eb7c
MAINT Remove support for the deprecated sequence of sequences
...
MAINT Remove sequence of sequence support from datasets
MAINT Remove return_indicator param
MAINT Remove multilabel-seq test in OVR
MAINT Remove multilable-seq test in check_cv
MAINT Remove multilabel seq test in label_binarizer
TST type_of_target returns "unknown" for multilabel-sequence types
TST _check_targets should raise a ValueError
DOC show multilabel indicator as an example; remove return_indicator param
DOC use consistent lower case y for target
2015-08-11 22:26:44 +05:30
a8626b36a6
TST/COSMIT remove nose call boilerplate
2015-05-28 14:54:01 -04:00
b9d8fd56a2
ENH support for sparse coef_ in ovr classifier
2015-05-15 20:26:38 -04:00
d89c215013
Add tags to classifiers and regressors to identify them as such.
2015-03-31 19:59:49 -04:00
cd2ee7e454
MAINT docstring --> comments to prevent nose from using doc in verbose mode
2015-03-21 11:16:49 +05:30
deeaf0ad35
change default to shuffle=True in SGDClassifier and friends.
2015-03-03 11:56:23 -05:00
e20d1a4391
ENH decision_function now returns #votes and scaled scores.
...
ENH use the decision_function to compute the prediction
TST Added test to assertain correlation of decision function and prediction
Add test to check if ties are broken using decision function
Add decision_function tests to check for votes/scores
ovo_ties test will use decision function to calculate votes and score.
DOC Add entry to OvO's doc regarding how ties are broken
DOC Clean up the predict / decision_function methods' docstring.
PEP8 Minor PEP8 clean up.
2015-02-09 14:39:11 +05:30
40c03d4880
COSMIT pep8 fixes to sklearn/tests
2014-12-18 16:40:01 +01:00
744614dfba
Check if targets is a numpy array and convert it into one if it isn't
2014-11-30 11:26:38 +00:00
b584ac47be
FIX ovr predict_proba in the binary case
...
Previous commit fixed the shape, but swapped the actual probabilities.
2014-09-29 22:51:19 +02:00
6e123c0661
Fixes ovr in the binary classifier case, and adds support for lists of feature arrays in the multiclass case.
2014-09-26 21:31:41 -04:00
0807e19dc2
MAINT deprecate fit_ovr, fit_ovo, fit_ecoc, predict_ovr, predict_ovo, predict_ecoc and predict_proba_ovr
2014-07-19 12:17:22 +02:00
eb2098f26f
Closes #2360 . Fix tiebreaking.
2014-07-18 16:48:08 +02:00
9f45e3730f
Modified sparse OvR to handle sparse target data
...
Defaulted label binarizer to set sparse_output=True when training ovr
classifiers, edited Label binarizer to allow for sparse binary column output
2014-07-18 12:10:50 +02:00
e072fbce62
Turn useless line of code into descriptive comment
2014-07-15 12:44:05 +02:00
e633f25233
ENH make_multilabel_classification for large n_features: faster and sparse output support
2014-07-08 11:13:58 +02:00
82864ad73b
FIX avoid using sequences of sequences and fix tests
2014-06-24 16:27:05 -04:00
d2db017058
TST use assert_warns and modernise test for constant predictor
2014-06-24 15:55:21 -04:00
51d4f75d06
FIX case where label is constantly absent
2014-06-23 11:53:46 -04:00
fa76ac38ee
FIX OvR with constant label for non-predict methods
2014-06-23 10:52:01 -04:00
28eb1ebe15
Assert or ignore all sequence of sequences deprecation warnings
2014-06-05 11:20:36 +02:00
8eeda198d2
[tests] Use Python 3 zip/Python 2 zip consistently in tests
2014-03-03 13:02:17 +00:00
0f5685b8ee
FIX predict_proba status on SGD and SVC when disabled
...
Using a property trick suggested by @mblondel. The following now
works:
>>> clf = SVC(probability=False)
>>> hasattr(clf, "predict_proba")
False
This simplifies the smoke tests.
Added a note to SVC stating that this should be done in new code.
2014-02-27 18:30:35 +01:00
3ee7b72426
BUG: OneVsOneClassifier was broken with string labels
2013-12-18 22:51:22 +01:00
e488704849
COSMIT pep8
2013-07-23 15:17:32 +02:00
784486126f
Cleaned up tests, and removed unused multilabel parameter in decision_function_ovr
2013-07-22 12:56:05 +02:00
baffa3017b
Strengthened tests for OneVsRestClassifier decision_function
2013-07-22 12:56:05 +02:00
5088875a0c
Updated fixes for #2012 .
2013-07-22 12:56:05 +02:00
cbd65449c0
Added decision_function support to OneVsRestClassifier and a test, test_ovr_single_label_decision_function, in test_multiclass.py
2013-07-22 12:56:05 +02:00
d33634d0ce
ENH add multilabel support to PRF metric family
...
Support also in classification_report
2013-05-20 13:08:50 +10:00
34022b5283
FIX scores calculation in ovo multiclass
...
Verify that ties can be won by all labels
2013-05-04 10:33:49 +02:00
5bcf25fc2f
ENH multilabel metrics: accuracy, Hamming, 0-1 loss
2013-03-02 20:41:34 +01:00
645216fb2f
Merge pull request #1524 from amueller/break_ovo_ties
...
MRG break ties in OvO using scores
2013-02-14 06:15:32 -08:00
541987802f
P3K range vs. xrange
2013-02-14 02:05:35 +01:00
0449d51029
TST test for breaking OVO ties
2013-02-04 22:43:39 +01:00
3be27ff640
Cosmits and typos.
2012-12-03 00:18:52 +09:00
482e57685f
Spelling is checked, passes Flake8 without errors.
2012-12-02 09:32:28 +00:00
48f0a44640
corrected bad test in test_multiclass
...
`test_ovr_single_label_predict_proba` wasn't checking consitency between
`predict_proba` and `predict` correctly. Now it is. Nose tests pass except for
1 conerning PIL.
2012-12-02 09:32:28 +00:00
c4cbffcc13
divided test for predict_proba into two functions
2012-12-02 09:32:27 +00:00
fd930e37da
wrote test_ovr_predic_proba method
...
If anyone has any suggestions for additional testing methods, it would be much
appreciated. Currently implemented:
- Do probabilities sum to one in single label case?
- Is a ValueError raised for base classifier with no predict_proba method?
- Do you arrive at the same predictions from predict_proba and predict?
2012-12-02 09:32:27 +00:00
c1ffbb47e7
Missing random_state in LinearSVC.
2012-11-03 00:53:34 +09:00
afb666c1d0
Use sklearn.utils.testing in tests.
2012-10-25 15:31:35 +09:00
10d9023fca
ENH make multi-class more robust in discovering scoring functions
2012-10-07 20:25:50 +02:00
408d3ef86d
COSMIT, removed scikits.learn things, removed orphan file.
2012-07-23 22:50:12 +01:00
d9226d0f9e
Merge pull request #886 from amueller/multiclass_always_present
...
MRG: enh in OvR, use constant predictor if one class always present or never.
2012-07-23 05:48:30 -07:00
a4845351af
ENH classes_ on OvR classifier
...
Getting to the classes via label_binarizer_ violates the Law of Demeter:
"Only talk to your immediate friends."
(https://en.wikipedia.org/wiki/Law_of_Demeter )
2012-07-20 18:29:35 +02:00
839c4b087a
MISC address Gael's and Lars's comments, make ECOC tests deterministic.
2012-07-15 14:04:00 +01:00
b463f8f2d0
ENH in OvR, use constant predictor if one class always present or never.
...
Closes #559 .
2012-07-15 14:04:00 +01:00
eeb6dcc4cc
ENH use random states everywhere, never call np.random.
2012-05-06 18:35:14 +02:00
Raghav R V