77b4670b6a
COSMIT use np.clip in SGD
2012-10-29 00:47:00 +01:00
53b220358c
COSMIT correct error msgs in SGD and make them more consistent
2012-10-14 12:09:59 +02:00
97ed66151a
BUG + DOC l1_ratio in SGD and CD
...
SGD{Classifier,Regressor} treated l1_ratio as previously rho. Docstrings
stated that l1_ratio must be >0, which means pure L2 regularization would
not be possible.
2012-10-14 11:43:26 +02:00
72e71bb439
ENH address @agramfort's comments, fix some doctests
2012-10-11 20:56:03 +01:00
b39609af4e
ENH rename rho in SGD
2012-10-11 20:56:03 +01:00
55ef1d617a
DOC: protect `classes_` for valid rst
2012-10-07 16:39:40 +02:00
77e69edcec
COSMIT rm deprecated stuff -- lots of it
2012-09-20 21:40:33 +02:00
1313f25cda
COSMIT pep8
2012-09-20 14:49:14 +01:00
dc943722de
COSMIT refactor linear classifiers
...
Factored out the common code to LinearSVC, LogisticRegression, Perceptron
and SGDClassifier into a LinearClassifierMixin.
2012-09-17 21:06:36 +02:00
f15e5803ef
COSMIT rm deprecated SGDClassifier.classes property
2012-09-16 22:18:36 +02:00
69176b11dd
DOC small fixes to SGD docstrings
2012-09-14 15:12:34 +02:00
5ecb9ac332
add predict_log_proba and test; better docstrings
2012-09-14 14:44:34 +02:00
9feb095149
fix docstring of predict_proba
2012-09-13 12:52:21 +02:00
5d357f7df1
followed @larsmans tip to get rid of _decision_function
2012-09-13 08:58:04 +02:00
523ba8fc27
cosmit
2012-09-12 18:50:24 +02:00
ad9cc064bf
fix the predict_proba w/ sparse matrix regression by using shape instead of len
2012-09-12 18:38:30 +02:00
60c730d21e
rename get_loss_function to _get_loss_function
2012-09-07 10:08:37 +02:00
6a23fc8165
use DEFAULT_EPSILON consistently
2012-09-06 22:07:26 +02:00
63608c67ae
refactored input validation; special loss function factory for huber and epsilon insensitive loss
2012-09-06 21:41:57 +02:00
dfd2140d59
move get_loss_function to _partial_fit
2012-09-06 10:04:49 +02:00
70099dd40d
rm instance variables learing_rate_type, loss_function, and penalty_type; create them before plain_fit
2012-09-06 09:35:09 +02:00
f524edcf89
DOC forgot dot in SGDCLassifier documentation
2012-09-04 07:46:16 +02:00
ab7e942233
DOC add support for sparse arrays to SGDCLassifer
2012-09-04 07:46:16 +02:00
5236f8c88f
DOC Fixed a return in predict_proba in SGDClassifier
2012-09-04 07:46:16 +02:00
bcbcc5cd87
DOC Added SGDCLassifier support only binary prediction probabilites.
2012-09-03 19:25:48 +02:00
81bca36f6b
fix typo
2012-09-03 10:33:56 +02:00
0441dc58b5
Merge branch 'master' into sgd-yshape-fix
2012-09-03 10:25:39 +02:00
9e7c03f71a
Add or fix deprecation schedule in warnings.
2012-09-01 20:02:14 +02:00
d9ff3f5a90
cosmit: constants for penalty types and learning rate types; inline comments;
2012-08-30 09:29:15 +02:00
a1dfd317aa
inline comment
2012-08-30 09:13:27 +02:00
e0434de8ef
_tocsr not needed because of atleast2d_or_csr
2012-08-30 09:12:46 +02:00
0902ed09a5
ravel y; use atleast2d_or_csr for input validation
2012-08-30 09:11:56 +02:00
c7e65d24e4
rm redundant input check (we check in _partial_fit)
2012-08-30 08:22:39 +02:00
55bc245eed
changed shape of predict_proba
2012-08-16 16:49:30 +02:00
913e2224a4
COSMIT minor refactoring of SGD
2012-08-02 18:43:17 +02:00
7c8f435469
COSMIT move BaseSGD to its only place of usage
2012-08-01 15:25:36 +02:00
1ebcc4a486
COSMIT pep8
2012-05-15 21:18:30 +02:00
611d1b5b00
Merge branch 'master' of github.com:scikit-learn/scikit-learn
2012-05-11 21:28:42 +09:00
85eccb9b39
Address @ogrisel and @pprett's comments.
2012-05-11 21:19:50 +09:00
4302abc286
Merge branch 'sgd_losses' of github.com:mblondel/scikit-learn into sgd_losses
...
Conflicts:
doc/whats_new.rst
sklearn/linear_model/tests/test_sgd.py
2012-05-11 21:13:09 +09:00
6592fdbee4
P3K: Fixing prints and dtypes
2012-05-11 11:24:27 +02:00
33006bc900
MISC removed unused lines, see #666 .
2012-05-05 18:57:09 +02:00
62f4c8ccb5
predict_proba with loss="modified_huber".
2012-05-03 16:53:47 +09:00
3e762a0d4a
Add epsilon-insensitive loss.
2012-05-02 18:08:11 +09:00
6204829e43
Allow regression losses for classification.
2012-05-02 17:37:23 +09:00
7b75227afb
Rename "p" to "espilon".
2012-05-02 17:23:21 +09:00
5d2d7d1d06
rm unused vars
2012-04-13 10:49:39 +02:00
b68e6bedb1
asarray not needed because of check_arrays stmt above
2012-04-13 10:42:53 +02:00
6ef572c1e5
fix issue 762: SGDRegressor does not clear coef_ from previous fit
...
consolidated _partial_fit signature of SGDClassifier and SGDRegressor.
Parameter memory is allocated in _partial_fit.
2012-04-13 10:40:51 +02:00
7fcb2dbb3e
moved WeightVector und SequentialDataset into seperate modules.
2012-03-07 20:09:05 +01:00
Lars Buitinck