a87011262a
FIX precision to float64 across the codebase
2015-10-16 11:20:29 +02:00
2f099334f9
MAINT deprecate 1d input arrays for all estimators
...
Passing 1D arrays to check_array, without setting `ensure_2d` to false now
raises a deprecation warning before reshaping it. This will later throw an
error.
All Scaler classes also throw warnings when 1D arrays are passed.
All unit tests/doctests are modified to ensure that no 1D arrays are passed,
except in explicit 1D array tests where the warnings have been silenced.
Additional tests are also included which check for different 1D array cases.
2D array tests with one samples and one features are also added and where
they failed, `check_array` call has been modified to give a more useful error
message
2015-09-09 15:49:58 +02:00
434ee958aa
test for accepted sparse matrices
2015-08-26 22:00:55 -04:00
0650d5502e
DOC adding backlinks to docstrings
2015-06-03 00:24:04 -04:00
8702801c13
used special.fdtrc instead of fprob
...
added integer division
2015-03-23 21:46:59 +05:30
2bc7ffef1c
Add "See also" for selectors and scoring funs
...
Details:
- Reference uni variate selectors and scoring funcs
- Indicate chi2 is only for non-negative features
- Add "See also" formatting rules to contributing
- Add example of good "References" in docstrings
- Misc small doc updates for uni feature selection
- Add missing periods (for consistency)
- Anova -> ANOVA (for consistency)
- Make a numbered list render per item per line (for readability)
- Capitalize the first letter of a numbered list (for readability)
2015-03-10 21:40:25 +08:00
95f1d9562b
FIX DOC TST: alpha is an uppper bound on FDR
2015-02-24 16:26:24 -05:00
700babc43f
extensive (excessive?) testing of FDR
2015-02-24 16:26:24 -05:00
1add39ff11
[Feature Selection] Fix SelectFDR thresholding bug ( #2771 )
...
From https://github.com/scikit-learn/scikit-learn/issues/2771 , we were
not correctly scaling the alpha
parameter (http://en.wikipedia.org/wiki/False_discovery_rate#Benjamini.E2.80.93Hochberg_procedure )
with the number of features (== hypothesis). Thus, the alpha
parameter was not invariant wrt the number of features.
The correction is as suggested in the original issue, and a test has
been added that verifies that for various numbers of features, an
appropriate false discovery rate is generated when using the selector.
2015-02-24 16:25:58 -05:00
bc5b403af3
FIX #4059 : explicit warning for strict selectors
2015-02-06 12:06:34 +01:00
abd31d2cad
MAINT Make uniform the error raised for not fitted condition
2015-01-12 00:33:40 +05:30
bee0a98a1e
cosmit in warning message
2014-10-08 21:04:59 +02:00
22cc25fc61
ENH: More descriptive error which prints the feature indices
2014-10-08 20:09:32 +02:00
0bd095f9d6
FIX: Raise warnings in f_classif a given feature is constant throughout
2014-10-08 19:32:47 +02:00
b3bdb08964
DOC fix formatting of attributes etc. in docstrings
2014-07-28 19:04:59 +10:00
c235c3b566
ENH add allowed_sparse named argument for @ogrisel
2014-07-20 15:31:28 +02:00
6e2a83b4e1
remove check_arrays stuff and old input validation
2014-07-20 13:31:45 +02:00
185beb3e78
DOC remove comment thanks to @jnothman
...
'All our feature selectors output masks now (as they should, IMO).
It will be turned back into indices only for sparse matrix indexing,
but only until we require scipy >= 0.14 (?).'
2014-05-08 20:56:36 +02:00
e58405d6f6
MAINT remove redundant class hierarchy + fix api perform parameter check in fit
2014-05-08 20:56:21 +02:00
a3f81d9a04
Merge pull request #3010 from griffinmyers/fix-select-k-best
...
Fixed SelectKBest corner case: k=0
2014-03-28 09:32:31 +11:00
f1e2e1c874
Fixed SelectKBest corner case: k=0
...
Without this fix k = 0 will have the same behavior as k = 'all' based on how
the array indexing was written.
I've included a test that demonstrates proper behaviour.
2014-03-27 15:18:50 -04:00
c1003ec51d
DOC heapsort is not stable at all
2014-03-27 11:19:29 +01:00
4d18fd8837
BF in f_regression: variable naming + use assert_*array*_almost_equal.
2014-02-07 19:28:06 +01:00
8e0b4be57e
BF: Correct degrees of freedom in f_regression + test.
2014-02-07 19:28:05 +01:00
f31d6d7538
pyflakes
2013-12-16 22:03:30 +01:00
44270ddc83
fix f_oneway with ints
2013-12-16 22:02:05 +01:00
551625fad7
MAINT simplify f_oneway in feature selection
...
reduce(lambda x, y: x + y, s) == sum(s), always.
2013-10-14 00:15:31 +02:00
9e10ec017b
COSMIT use norm function in feature selection
2013-10-13 23:20:39 +02:00
e8bad3d24a
DOC ASCII only in docstrings
...
As discussed on the mailing list.
2013-08-28 18:22:57 +02:00
b46ea718af
FIX remove warnings from univariate FS
...
These warnings are issued practically always when using
frequency-valued or boolean data.
Switched to a stable sort to get reproducible results.
2013-08-19 14:56:37 +02:00
bc2acb519c
FIX yet another chi2 bug
2013-06-01 17:42:04 +02:00
b486875497
FIX bug (swapped args) in chi2
...
Added a non-regression test.
2013-06-01 17:26:20 +02:00
194fa350d8
ENH speed up sklearn.feature_selection.chi2
...
New version is ~4× faster and reuses allocated arrays.
2013-05-30 12:23:29 +02:00
9a8845e913
ENH Create FeatureSelectionMixin for shared [inverse_]transform code
...
Also rename FeatureSelectionMixin -> SelectorMixin -> _LearntSelectorMixin
And rename sklearn.feature_selection.{selector_mixin -> from_model}
2013-05-21 13:53:12 +10:00
489d294a87
ENH Feature selection should use CSC matrices
2013-05-07 22:31:23 +10:00
2b57e4e52e
Merge pull request #1935 from jnothman/base_filter_inv_transform
...
Small improvements to inverse_transform used in feature_selection.
2013-05-07 05:26:58 -07:00
1b07773ca7
DOC comment feature selection sparse inverse_transform
2013-05-07 16:00:06 +10:00
4fd975ce80
ENH sparse _BaseFilter.inverse_transform
2013-05-07 16:00:05 +10:00
14f22a91bb
DOC Fix comment for _BaseFilter.inverse_transform
2013-05-07 16:00:05 +10:00
313193fb0f
ENH _BaseFilter.inverse_transform should respect dtype
2013-05-07 16:00:04 +10:00
0480a66814
FIX ensure SelectPercentile only removes tied features in case of ties
2013-05-07 12:34:47 +10:00
7af4fd934e
ENH call asarray on returned scores and pvalues
2013-05-07 12:15:11 +10:00
2ffb484adb
COSMIT pep8
2013-05-04 10:43:39 +02:00
eae18a3784
PY3 fix metaclasses. See #1829 .
2013-04-04 01:07:20 +06:00
b80f6a80b0
ENH allow SelectKBest to select all features in a parameter search
2013-03-13 12:28:49 +01:00
1967a0b323
P3K use six to have a python 2 & 3 compatible code base
2013-02-09 18:08:44 +01:00
19ed9b0165
ENH refactor univariate feature selection
...
Gets rid of "duplicate p-values" warning with filters that don't
actually use p-values.
2013-02-03 20:04:23 +01:00
9e3ef3985e
DOC some sphinx fixes
2013-01-16 22:51:24 +01:00
69210dabfd
COSMIT doc formating
2013-01-06 15:24:58 +01:00
754c2d9533
COSMIT pep8
2012-12-22 16:48:17 +01:00
Raghav R V