e304597fab
Raise error if file doesn't exist.
2011-06-21 17:57:21 +09:00
5bdc3ab5e6
Parallel vectorizing is slower than serial
...
Updated comments in feature_extraction.text.
This is unfortunate, as vectorizing is the bottleneck in text
classification, esp. with fast learners such as Naive Bayes.
2011-06-20 11:37:44 +02:00
4b172d11a0
Add note to base fit_transform doc.
2011-06-20 16:16:57 +09:00
fcb81abf6f
BUG: LassoLARS: X: modified during the normalization
2011-06-19 20:15:44 +02:00
e65e93e5ca
Add test and cleanup for 2c1c88
2011-06-19 15:16:34 +02:00
2c1c88b6e7
BUG: change alpha scaling in LassoLARS
2011-06-18 19:12:30 +02:00
45f38b28ed
Merge pull request #216 from ametaireau/master
...
minor doc typos
2011-06-18 08:01:44 -07:00
e5d19f8b35
BUG: Fix bug introduced in 83cf11c
2011-06-18 15:38:43 +02:00
cb43d6505e
Merge branch 'master' of https://github.com/scikit-learn/scikit-learn
2011-06-18 15:04:31 +02:00
4d3926b7b6
fix restructured text problems in the developers doc
2011-06-18 15:00:07 +02:00
c312618bdd
fix a typo in neighbors docs
2011-06-18 14:59:41 +02:00
60b2597ea9
ENH: Add a ShuffleSplit cross-validation iterator
2011-06-18 14:30:20 +02:00
83cf11c398
COSMIT: Minor refactor in lars_path
...
This will make LarsCV easier
2011-06-18 10:34:35 +02:00
6f8bd76ced
ENH: LARS: fit_params -> class params
2011-06-18 10:16:48 +02:00
f2f0caef27
DOC: Tweak the cross-val lasso path example
2011-06-18 09:33:14 +02:00
d4982ef199
ENH: Lasso and LassoCV: fit params -> class params
2011-06-18 09:21:27 +02:00
e61a052143
fit_transform does NOT return self + other docfixes
...
Removed some tutorial-style stuff and implementation details
WITHOUT putting those in their proper place (!).
2011-06-18 00:24:01 +02:00
a455d4d4b5
Documentation fixes.
2011-06-15 18:22:52 +09:00
668094e7c2
document placement new in SVMlight reader
2011-06-15 09:27:33 +02:00
dbaea6f177
typo
2011-06-15 00:58:23 +02:00
159bcb1dc4
FIX: make the dataset doctest fixture modular
2011-06-15 00:55:35 +02:00
b730d3a467
Fix broken doctest in mldata
2011-06-14 17:40:43 +02:00
08b7f87f5e
Python 2.5 and SciPy 0.7 (tentative) compat in mldata
...
Tested with SciPy 0.7.2.
2011-06-14 16:25:35 +02:00
20d7191bfc
Merge pull request #209 from mblondel/svmlight_format
...
Fast loader for the svmlight / libsvm format
2011-06-14 05:36:36 -07:00
483b3acad0
datasets/mldata.py is not a script, chmod 644
2011-06-14 11:10:16 +02:00
a7c75a2762
Merge branch 'master' into svmlight_format
...
Conflicts:
doc/modules/datasets.rst
2011-06-14 17:51:52 +09:00
535e910384
DOC: reorg of dataset page to make it more consistent
2011-06-14 10:25:49 +02:00
26cf261015
Merge branch 'master' into pberkes-mldata
2011-06-14 09:57:08 +02:00
f70c442f37
load_svmlight_format -> load_svmlight_file.
2011-06-13 22:31:23 +09:00
4f059a58fc
Merge pull request #7 from larsmans/mblondel-svmlight
...
Memory leak and exception safety in SVMlight reader
2011-06-13 04:34:05 -07:00
772b24a1ab
(Hopefully) full exception safety in SVMlight reader
...
Also some minor refactoring
2011-06-13 13:16:18 +02:00
2cede177f5
SVMlight reader: memory leak, type test
2011-06-13 12:56:29 +02:00
841a295ffc
Merge pull request #210 from larsmans/bernoulli-naive-bayes
...
Naive Bayes classifier for multivariate Bernoulli (binary features)
2011-06-13 02:50:56 -07:00
c92d52cf9e
Glitches in BernoulliNB and DiscreteNB (mostly docs)
...
No link for reference to Manning et al. since nlp.stanford.edu seems
to be down. Should add this later.
2011-06-12 21:50:04 +02:00
d59a06bc43
More comments.
2011-06-13 01:28:21 +09:00
446687a8bb
Fix ref counting bug.
2011-06-13 01:03:48 +09:00
650c464495
Fix compile issues on Mac OS X.
2011-06-13 00:54:29 +09:00
1a1510acbd
Merge branch 'mblondel-svmlight' of git://github.com/larsmans/scikit-learn into svmlight_format
2011-06-13 00:10:02 +09:00
c08faad239
BUG: broken doctest in BernoulliNB
2011-06-12 12:13:44 +02:00
5ab1feee4b
Update BernoulliNB docs
2011-06-11 19:52:12 +02:00
ea11518e0d
Extend MultinomialNB tests to BernoulliNB
...
All tests pass.
2011-06-11 19:44:37 +02:00
fb3ecab261
finish exception handling in SVMlight reader
...
Changes behavior: trying to open a non-existent file will result in
an IOError rather than a ValueError.
Slapped my name onto scikits/learn/datasets/svmlight_format.py out
of sheer vanity.
svmlight_format.py is now pep8 and pyflakes-clean.
2011-06-11 19:15:48 +02:00
8d5e00d975
Merge pull request #6 from larsmans/mblondel-svmlight
...
SVMlight reader update
2011-06-11 09:08:16 -07:00
6566057aa5
Use C++ exception handling in SVMlight reader.
...
Info from exceptions should be propagated up to the Python level,
but currently isn't.
2011-06-11 17:51:57 +02:00
24619a631d
Simplify and document SVMlight/libSVM data reader
2011-06-11 16:55:46 +02:00
20c15eef6c
SVMlight reader: skip one level of indirection
...
Uses placement new and vector::swap to eliminate some memory
allocations. Main benefits: exception safety will be easier to
guarantee and there are fewer opportunities for memory leaks.
Still, this is a hairy beast.
Also some other safety improvements.
2011-06-11 14:21:49 +02:00
d7debebe0c
SVMlight reader: cosmetic
2011-06-11 13:27:57 +02:00
a1156dafd4
SVMlight reader: one more clear() instead of delete
2011-06-11 12:46:40 +02:00
8526ab686c
Plumb memory leak in SVMlight reader
...
Quick fix, maybe only partial: the deallocator seems not to be called
in unit testing.
2011-06-11 12:35:17 +02:00
7eab00f46f
Fix comment.
2011-06-11 13:54:12 +09:00
Mathieu Blondel