Commit by debian on yoh/master :: r151df8c pymvpa/mvpa/ (2 files in 2 dirs):

Merge remote branch 'alioth/yoh/master'

• alioth/yoh/master: BF: moving scipy-dependent statement in gnbsearchlight BF+RF: apparently epsilon is tube_epsilon for sg.LibSVR + sg.svm.seed

Commit by debian on yoh/master :: rce6245b pymvpa/doc/ (2 files in 2 dirs):

Merge remote branch 'alioth/mh/master' into yoh/master

• alioth/mh/master: DOC: Restructure paper listing. DOC: Note new users with Nature Neuroscience paper ;-)

Commit by debian :: r151df8c pymvpa/mvpa/ (2 files in 2 dirs):

Merge remote branch 'alioth/yoh/master'

• alioth/yoh/master: BF: moving scipy-dependent statement in gnbsearchlight BF+RF: apparently epsilon is tube_epsilon for sg.LibSVR + sg.svm.seed

Commit by debian :: rce6245b pymvpa/doc/ (2 files in 2 dirs):

Merge remote branch 'alioth/mh/master' into yoh/master

• alioth/mh/master: DOC: Restructure paper listing. DOC: Note new users with Nature Neuroscience paper ;-)

Commit by debian :: rbbe31f1 pymvpa/mvpa/measures/gnbsearchlight.py:

BF: moving scipy-dependent statement in gnbsearchlight

Commit by debian :: r9eca119 pymvpa/mvpa/clfs/sg/svm.py:

BF+RF: apparently epsilon is tube_epsilon for sg.LibSVR + sg.svm.seed

Commit by debian :: r1d1556f pymvpa/ (11 files in 8 dirs):

Merge commit 'alioth/yoh/master' into yoh/master

• commit 'alioth/yoh/master': NF: mvpa.support.ipython + ipy_pymvpa_completer BF+DOC: Wrong link to tutorial. BF: actually fail (error exist status) whenever unittest-debug fails (uff) BF: access selected_ids from .ca BF: unittest-debug -- do not rely on MVPA_SEED or OK being in the tail -- just go to the end ;) BF: nr_class can be None, so just use %s in debug printout OPT+PL: more efficient loop while extracting info from /proc/pid/status BF: access ca.confusion instead of just .confusion DOC(BF): forgotten import to make snippet work

Commit by debian :: rf3e2ba4 pymvpa/mvpa/clfs/sg/svm.py:

BF: workaround for shogun -- epsilon in constructor for regressions has no effect

Commit by debian :: rc534630 pymvpa/ (4 files in 3 dirs):

 NF: mvpa.support.ipython + ipy_pymvpa_completer

that is where to collect all additions for making IPython experience
most pleasant ;)

Commit by debian :: ref700df pymvpa/mvpa/kernels/sg.py:

BF: import debug for kernels.sg

Commit by debian :: r63743d1 pymvpa/mvpa/misc/errorfx.py:

ENH: complement CorrErrorFx with just CorrFx

Commit by debian :: r0711b0a pymvpa/mvpa/ (atlases/fsl.py tests/test_atlases.py):

ENH: fslatlas.get_maps -- accepts now 'key_attr' defining on what to use for keys

Commit by debian :: r885bb8d pymvpa/mvpa/ (3 files in 3 dirs):

ENH: more debugging for sg kernels and svm

Commit by debian :: r50a7822 pymvpa/ (5 files in 4 dirs):

RF: atlas.levels now corresponds to previous atlas.levels_dict

Commit by debian :: r8d2d7ec pymvpa/bin/atlaslabeler:

 RF/ENH: atlaslabeler -- ability to provide a stream of voxels indices

also updated for last RF to obliterate query_voxel argument

whole thing is still a mess -- need some though to refactor it properly

Commit by debian :: r51a7470 pymvpa/mvpa/ (atlases/base.py tests/test_atlases.py):

RF: atlases __getitem__ for queries by voxel indices, __call__ -- by coordinates

Commit by debian :: r349d467 pymvpa/ (6 files in 5 dirs):

RF: for the sake of consistency -- n_levels -> nlevels (atlases, wavelet mapper)

Commit by debian :: r0e3ab88 pymvpa/mvpa/measures/gnbsearchlight.py:

 BW: make gnbsearchlight compatible with scipy.sparse as of 0.6.0

API was changed later on to use shape instead of dims

Commit by debian :: r93e86ab pymvpa/mvpa/clfs/warehouse.py:

 BW: do not instantiate krr for older shogun versions -- API is a bit different from the rest

just more pain than use -- upgrade shogun to >= 0.9.1

Commit by debian :: r5c8a2e7 pymvpa/mvpa/clfs/sg/svm.py:

 BF: no need to double seed shogun -- it wasn't the problem

0.9.1-2 has a jump on uninitialized value -- that seems to be the
cause of the mess.