6cf1f6abba
DOC GaussianProcess (classification and regression) examples added to docstrings ( #11748 )
2018-08-18 19:55:04 +03:00
Hanmin Qin
9b8fd0b0d0
DOC Correct default n_jobs & reference the glossary ( #11808 )
...
Also improves the glossary entry for n_jobs.
2018-08-18 19:57:10 +10:00
40e6c43cb4
Joblib 0.12.2 ( #11741 )
...
* joblib 0.12.2
* Export _joblib's register_parallel_backend
* Use latest version of coverage
2018-08-03 12:34:25 +02:00
eeee6a5085
MNT Replacing log with log1p where applicable. ( #11428 )
2018-07-14 21:05:03 +08:00
58f84694a8
DOC Add reference to the glossary in the warm_start parameter description of public APIs ( #10699 )
2018-02-26 20:21:17 +08:00
7700b5ac73
[MRG+1] FIX: Use ConvergenceWarning whenver it applies ( #10306 )
2018-02-10 05:16:08 -08:00
45dc891c96
[MRG+2] Clean common tests ( #9340 )
...
* rm dupes
* add check_supervised_y_no_nan in classifier checks: this implies changes for Ridge classifiers
* fix docstrings/comments
* FIX check fitting 1d X array raises error and FIX check fitting 2d array with only 1 feature either works or returns informative message
* modify check_fit2d_1sample in common tests so that it checks fitting either works or returns an informative message
* rm SpectralClustering case for the moment
* uniformize error messages for 1 sample case and fix SpectralClustering with ensure_min_samples=2
* add unit test for mean_shift when n_samples * quantile < 1
* FIX travis with ensure_min_samples=2 in _PLS
* try fix for failing tSNE test
* typos
* take @agramfort's review into account
* sc to fix string in gaussian_process
* add the class that is present to preserve information of previous message in gpc.py
2017-09-26 18:13:15 +02:00
a47c3b9a07
ENH use expit in ``_BinaryGaussianProcessClassifierLaplace`` ( #9011 )
...
SciPy already provides a numerically stable ``special.expit``
function, which does not overflow on the example mentioned in #8641 .
2017-06-07 10:20:54 +10:00
e3c9ae204f
[MRG+1] DOC improve description and consistency of random_state ( #8689 )
...
* DOC improve description of random_state in train_test_split
* DOC Make random_state consistent through documentation
* FIX reverse doc mistake
* FIX address comment of Tom
* DOC address comments
* DOC remove empty line
* DOC remove unecessary white spaces
2017-04-05 17:43:21 -07:00
8570622a44
[MRG+1] DOC insert spaces before colons in parameter lists ( #7920 )
...
* DOC insert spaces before colons in parameter lists
Complies with numpydoc to improve rendering and automatic quality
assurance such as #7793 . Affects listings of Parameters Attributes,
Returns.
Performed with the help of:
grep -nE '^( )+[a-zA-Z][a-zA-Z0-9_]*: ' sklearn -R | grep -v -e
externals -e tests | grep -v -e default: -e else: -e Warning: -e Note:
-e TRAIN: -e Default: -e True: -e False: -e DOI: -e In: | gsed
's|\([^:]*\):\([0-9]*\):\([^:]*\):\(.*\)|--- a/\1\n+++ b/\1\n@@ -\2,1
+\2,1 @@\n-\3:\4\n+\3 :\4|' | git apply --unidiff-zero -
* DOC fix numpydoc format for param
2016-11-25 10:59:22 +01:00
9551ec85ea
DOC add spaces before colons in docstrings ( #7589 )
2016-10-06 14:01:12 +11:00
60a1356f31
[MRG + 1] More versionadded everywhere! ( #7403 )
...
* insert versionadded versionchanged directives in docstrings for 0.18
indicate where exception classes were moved from
* moved versionadded in the proper places
2016-09-27 16:19:47 -04:00
0485ada58b
General spelling fixes
2015-12-16 09:46:42 -08:00
2653833a07
DOC some fixes to the doc build.
2015-11-03 12:23:23 -05:00
2649598c64
FIX Enforcing y to be numeric in GPR and fixing python3 issue with maps
2015-10-19 18:07:14 +02:00
3957fc3b34
ENH GPR.log_marginal_likelihood() returns the current log-likelihood if no theta vector is provided
2015-10-19 18:07:14 +02:00
8f77ced609
MISC Changing default for copy_X_train to True in GaussianProcessClassifier
2015-10-19 18:07:14 +02:00
e2a340a4ee
REF More meaningful names for hyperparameters of GP kernels
2015-10-19 18:07:14 +02:00
2dd1f285b9
PEP8 Making pep8 and pyflakes happy
2015-10-19 18:07:14 +02:00
1c4b23f47b
MISC copy_X_train defaults to True in GPR and GPC
2015-10-19 18:07:14 +02:00
e87fad31fb
REF Renaming max_iter to max_iter_predict in GaussianProcessClassifier
2015-10-19 18:07:14 +02:00
c6db24e56f
REF Several small changes based on @eickenberg's suggestions
2015-10-19 18:07:13 +02:00
d34b7ddbfc
REF Realizing multi-class GPC bei delegating to OneVsRest or OneVsOne
...
Delegation is more flexible than subclassing.
2015-10-19 18:07:13 +02:00
34506ad705
COS Making BinaryGaussianProcessClassifierLaplace private
2015-10-19 18:07:13 +02:00
032c2a763b
FIX Added missing import of ConstantKernel in gpc.py
2015-10-19 18:07:13 +02:00
83bc6c2440
ENH Adding max_iter parameter for Newton iteration in GPC
2015-10-19 18:07:13 +02:00
c57aa64312
ENH X_train_ can be copied if requested
2015-10-19 18:07:13 +02:00
adc7451201
PEP8 Fixing PEP8 issues in gaussian_process package
2015-10-19 18:07:13 +02:00
59d754ba90
FIX Hiding method decision_function in GaussianProcessClassifier
2015-10-19 18:07:13 +02:00
768c81fd1b
TST Further GP-related fixes
2015-10-19 18:07:13 +02:00
92a6354aaf
TST Fix GPC and GPR such that all tests are ok
2015-10-19 18:07:13 +02:00
27b30166c3
FIX Several minor changes/fixes in gaussian_process module
2015-10-19 18:07:12 +02:00
3a312625f5
ENH Support for multi-class GPC added based on OneVsRestClassifier
2015-10-19 18:07:12 +02:00
07cd4d4e05
ENH GPC can use an externally defined optimizer for hyperparameter tuning
2015-10-19 18:07:12 +02:00
bcd4b72897
REFACTOR Remove theta_ attribute from GPR and GPC
...
It can be accessed instead via kernel_.theta
2015-10-19 18:07:11 +02:00
4db5142bb8
REFACTOR GP kernel hyperparameters theta and their bounds are log-tranformed
...
This representation of the search space is more amenable for
hyperparameter search, as hyperparameters like
length-scales naturally live on a log-scale.
2015-10-19 18:07:11 +02:00
22f049b6bc
ENH GaussianProcessClassifier supports multiple restarts of the optimizer
2015-10-19 18:07:11 +02:00
caa4196c3a
DOC Revised docstrings of kernels.py
2015-10-19 18:07:10 +02:00
7773707e0a
DOC gpc module fully documented
2015-10-19 18:07:10 +02:00
1c70e52dd6
ENH Adding diag(X) method to GP kernels.
...
Using this method rather than np.diag(kernel(X)) or np.apply_along_axis
reduces memory-consumption and speeds up computation considerably
2015-10-19 18:04:56 +02:00
fb4eded5ee
ENH GPC can use warm-starts in _posterior_mode()
2015-10-19 18:04:56 +02:00
fd2cfc868c
ENH Reducing memory-consumption of GPC.predict_proba()
...
We compute the diagonal of var_f_star directly and not the whole matrix.
2015-10-19 18:04:56 +02:00
3043a126a4
FIX GPC uses LabelEncoder in fit() and returns (n_samples, n_classes) array in predict_proba()
2015-10-19 18:04:56 +02:00
e3423b989a
FIX Minor bugfixes and downward-compatibility/python3 issue
2015-10-19 18:04:56 +02:00
2feb6e02cf
REFACTOR Replacing all calls to deepcopy() by clone()
2015-10-19 18:04:55 +02:00
403b5f1412
REFACTOR REnamed kernel property params to theta
2015-10-19 18:04:55 +02:00
d0bc5dc35c
FIX GPs do not modify kernel attribute but store optimized kernels in kernel_
2015-10-19 18:04:55 +02:00
d8405c2403
REFACTOR Hyperparameter optimization in GPs enabled explicitly by optimizer parameter
2015-10-19 18:04:55 +02:00
e05daaec52
ENH Avoid to create square matrix W_sr in all but one place (TODO)
2015-10-19 18:04:55 +02:00
e165fed1cc
REFACTOR GaussianProcessClassification/Regression -> ...Classifier/Regressor
2015-10-19 18:04:55 +02:00
Adrin Jalali
Hanmin Qin