[pymvpa] Use case for recursive feature elimination
Michael Waskom
mwaskom at MIT.EDU
Mon Nov 1 21:45:53 UTC 2010
Hmm, on second thought, I can't quite seem to get this to work. Trying the
example from __test_matthias_question line for line, it throws a TypeError
on classifier initialization.
TypeError: Unexpected keyword argument transformer=<ufunc 'absolute'> for
BoostedClassifierSensitivityAnalyzer: 4 ca:ca{base_sensitivities raw_results
null_t null_prob+}. Valid parameters are ['base_sensitivities',
'raw_results', 'null_t', 'null_prob']
wtf, right?
Current date: 2010-11-01 17:43
PyMVPA:
Version: 0.5.0.dev
Path:
/software/python/nipype0.3/lib/python2.6/site-packages/pymvpa-0.5.0.dev-py2.6-linux-x86_64.egg/mvpa/__init__.pyc
Version control (GIT):
GIT information could not be obtained due
"/software/python/nipype0.3/lib/python2.6/site-packages/pymvpa-0.5.0.dev-py2.6-linux-x86_64.egg/mvpa/..
is not under GIT"
SYSTEM:
OS: posix Linux 2.6.31-22-generic #65-Ubuntu SMP Thu Sep 16
16:21:34 UTC 2010
Distribution: Ubuntu/9.10/karmic
EXTERNALS:
Present: cPickle, ctypes, good scipy.stats.rdist, good
scipy.stats.rv_discrete.ppf, griddata, gzip, libsvm, libsvm verbosity
control, lxml, matplotlib, nibabel, nose, numpy, pylab, pylab plottable,
pywt, pywt wp reconstruct, running ipython env, scipy
Absent: atlas_fsl, atlas_pymvpa, elasticnet, glmnet, h5py, hcluster,
lars, mdp, mdp ge 2.4, nifti, nifti ge 0.20090205.1, nipy, openopt,
pprocess, pywt wp reconstruct fixed, reportlab, rpy2, sg ge 0.6.4, sg ge
0.6.5, sg_fixedcachesize, shogun, shogun.krr, shogun.lightsvm, shogun.mpd,
shogun.svrlight, weave
Versions of critical externals:
ipython : 0.10.1
scipy : 0.7.2
numpy : 1.4.1
ctypes : 1.1.0
matplotlib : 1.0.0
lxml : 2.3.beta1
numpy : 1.4.1
pywt : 0.2.0
Matplotlib backend: TkAgg
RUNTIME:
PyMVPA Environment Variables:
PyMVPA Runtime Configuration:
[externals]
have griddata = yes
have pprocess = no
have good scipy.stats.rdist = yes
have pylab plottable = yes
have pywt wp reconstruct = yes
have mdp = no
have lxml = yes
have running ipython env = yes
have nibabel = yes
have sg_fixedcachesize = no
have elasticnet = no
have shogun.mpd = no
have matplotlib = yes
have pywt wp reconstruct fixed = no
have scipy = yes
have reportlab = no
have openopt = no
have libsvm = yes
have h5py = no
have shogun.krr = no
have nifti ge 0.20090205.1 = no
have nose = yes
have weave = no
have atlas_fsl = no
have ctypes = yes
have hcluster = no
have sg ge 0.6.4 = no
have sg ge 0.6.5 = no
have good scipy.stats.rv_discrete.ppf = yes
have libsvm verbosity control = yes
have mdp ge 2.4 = no
have shogun.svrlight = no
have shogun = no
have nipy = no
have glmnet = no
have lars = no
have nifti = no
have atlas_pymvpa = no
have cpickle = yes
have numpy = yes
have pylab = yes
have rpy2 = no
have shogun.lightsvm = no
have pywt = yes
have gzip = yes
[general]
verbose = 1
Process Information:
Name: ipython
State: R (running)
Tgid: 18461
Pid: 18461
PPid: 12889
TracerPid: 0
Uid: 85302 85302 85302 85302
Gid: 33368 33368 33368 33368
FDSize: 256
Groups: 1007 1008 33368 33376 68218 68236 68252 777333
VmPeak: 372540 kB
VmSize: 372512 kB
VmLck: 0 kB
VmHWM: 64688 kB
VmRSS: 64684 kB
VmData: 58408 kB
VmStk: 244 kB
VmExe: 2108 kB
VmLib: 53604 kB
VmPTE: 700 kB
Threads: 2
SigQ: 0/16382
SigPnd: 0000000000000000
ShdPnd: 0000000000000000
SigBlk: 0000000000000000
SigIgn: 0000000001001000
SigCgt: 0000000180000002
CapInh: 0000000000000000
CapPrm: 0000000000000000
CapEff: 0000000000000000
CapBnd: ffffffffffffffff
Cpus_allowed: ff
Cpus_allowed_list: 0-7
Mems_allowed: 00000000,00000001
Mems_allowed_list: 0
voluntary_ctxt_switches: 2295
nonvoluntary_ctxt_switches: 33
On Mon, Nov 1, 2010 at 2:10 PM, Michael Waskom <mwaskom at mit.edu> wrote:
> Hi Yarik,
>
> Thanks! The pointer to that test is very helpful indeed.
>
> Best,
> Mike
>
>
> On Mon, Nov 1, 2010 at 2:02 PM, Yaroslav Halchenko <yoh at dartmouth.edu>wrote:
>
>> Hi Mike,
>>
>> sorry that it causes you troble... but RFE is indeed somewhat difficult
>> to figure out...
>>
>> but before anything:
>> have you seen
>> http://dev.pymvpa.org/featsel.html#recursive-feature-elimination
>>
>> also look may be at
>> mvpa/tests/test_rfe.py: def __test_matthias_question(self):
>> which has one of RFE structures
>>
>> in general classifier with RFE as feature selection should be treated as
>> any other classifier
>>
>> also -- do you really think that RFE is necessary in your case? may be
>> some simpler feature selection or SMLR would do?
>>
>> P.S. it would be of greater benefit to everyone if we continue this
>> communication on the mailing list, if you don't mind of cause
>>
>> On Mon, 01 Nov 2010, Michael Waskom wrote:
>>
>> > Hey Yarik,
>>
>> > I'm trying to figure out how to use the RFE class in pymvpa 0.5dev.
>> > Sadly, there doesn't seem to be much in the (otherwise
>> > super-helpful) documentation about it. I get all of the parameters,
>> > but I'm not quite sure how it might fit into a typical analysis, and
>> > I'm very interested in trying out some recursive feature selection.
>> Do
>> > you have a simple use case you could put up as a gist (or elsewhere)
>> so
>> > I can get the basic idea? Nothing too didactic is required.
>>
>> > Thanks!
>>
>> > Mike
>>
>> --
>> Yaroslav O. Halchenko
>> Postdoctoral Fellow, Department of Psychological and Brain Sciences
>> Dartmouth College, 419 Moore Hall, Hinman Box 6207, Hanover, NH 03755
>> Phone: +1 (603) 646-9834 Fax: +1 (603) 646-1419
>> WWW: http://www.linkedin.com/in/yarik
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.alioth.debian.org/pipermail/pkg-exppsy-pymvpa/attachments/20101101/123f7ebe/attachment-0001.htm>
More information about the Pkg-ExpPsy-PyMVPA
mailing list