Part of #11992.
These were all the things that seemed pretty straight-forward. It's actually a bit bulky but should still be easy to review, hopefully.
* 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
* If self.kernel_ is None then self.kernel_ is set to RBFKernel in predict() as fit(). Fixes#6573
* xxx_ attributes can not be altered outside fit. Removing self.kernel_ from predict() and using previously implemented self.kernel attribute.
* Cleanup and add non-regression-test
* FIX: better error message for GPR
* TST: added tests for error message
* FIX: correct error message and simplified tests
* DOC: updated docstring for GPR
* FIX: pass float as value for alpha
* Raise original error after modification to preserve traceback
Amend test accordingly
* Make tests runnable with pytest without error.
Errors were due to pytest quirks with (deprecated) yield support.
* Add pytest build on Travis
and tweak pytest settings in setup.cfg
* Tweak comment
ensure that estimators only add private attributes and attributes with
trailing _
in cases when existing estimators don't follow this new rule, we deprecate the
attributes and make them follow this rule
* Remove Python 2.6 support
Some details about some slightly orthogonal changes:
* Note about cheking safely for nan is likely not valid any more (commit
introducing it is c80ca91b)
* scipy.linalg.qr econ parameter removed since scipy 0.9 in favour of
mode='economic'
* Remove unnecessary libgfortran in conda create command
* Putative fix by setting the random seed
* Revert unintended change
* Reinstate previous logic for checking for NaNs
* Reinstate change in error message
Error messages from Python 2.7 assertRegexp does not contain the
function name, in contrast with Python 3 assertRegex
* DOC framework for keeping API refs for deprecated classes/funcs
* DOC tagging deprecated for 0.20
* suggestion for LDA/QDA deprecation
* simplify deprecation message for GaussianProcess
* simplify deprecation messages
* fixup avoid import QuadraticDiscriminantAnalysis in qda.QDA (and similar for LDA).
* fixup test alias lda.LDA is instance of LinearDiscriminantAnalysis
* use less nose.tools import to simplify future transition to activly developing test suites/runners
* assert_equal -> assert_array_equal in test_feature_hasher_pairs_with_string_values
and one missed ImportError that should be replaced with AttributeError
* test for py2.6 compat with except AttributeError
* fix importing of SkipTest
* force using nose in python2.6 for now
* there was no assert_dict_equal in py2.6. but we can use assert_equal
although failed test will look a little bit ugly
* remove nose imports from doc/datasets
* insert versionadded versionchanged directives in docstrings for 0.18
indicate where exception classes were moved from
* moved versionadded in the proper places
Rüdiger Busche
Hanmin Qin