[pymvpa] >what does 405X 42696 in the dataset mean?

Yaroslav Halchenko debian at onerussian.com
Mon Dec 29 17:23:42 UTC 2008


>    it looks like dataset only shows a certain set of samples out of 925
>    after running feature_selection classifier.
FeatureSelection classifier should not alter number of samples!!! it
should just alter number of features.

>    If I just run the first loading part, I can see the intact 925 rows,
>    but after running it, it only shows 405 column.
BUT probably you've ran it within some CrossValidatedTransferError, or
SplitClassifier, SO, it was trained on a part of dataset. But dataset
itself should not get altered... what you are looking at?
clf.trained_dataset may be?

once again --if you prrovided complete script - it would have made it
easier to see what is happening


>    Thanks a lot for helping me out, Michael and Yaroslav.



>    Thanks,



>    On Mon, Dec 29, 2008 at 11:53 AM, Yune Lee <[1]life1994 at gmail.com>
>    wrote:

>    Hi,Yaroslav

>    That's a good point. I have no idea either.

>    Enclosed is the output by typing data dataset.summary()



>    Thanks,

>    YSL


>    On Mon, Dec 29, 2008 at 11:39 AM, Yaroslav Halchenko
>    <[2]debian at onerussian.com> wrote:

>      so bold file does have 925 volumes,
>      thus NiftiDataset I bet has also 925 samples after you load it
>      we can only wonder where 925-405=420 volumes got lost ;)
>      my guess could be you did some samples selection by the labels?
>      smth like
>      ds['labels', [1,2,3]]
>      ?
>      if not -- then just run and post output for

>    attr = SampleAttributes('post_AvB_onsets.txt')
>    dataset = NiftiDataset(samples='wrbold_25jun08MG.nii.gz',
>     labels=attr.labels, chunks=attr.chunks, mask='mask_25jun08MG.nii.gz')

>      print dataset.summary()
>      It Must have 925 samples ;-)

>    On Mon, 29 Dec 2008, Yune Lee wrote:
>    >    Many thanks for letting me know that the firt two lines are
>    redundent.
>    >    I'll later on just type from mvpa.suite import * only.
>    >    Any way, enclosed is the output.
-- 
Yaroslav Halchenko
Research Assistant, Psychology Department, Rutgers-Newark
Student  Ph.D. @ CS Dept. NJIT
Office: (973) 353-1412 | FWD: 82823 | Fax: (973) 353-1171
        101 Warren Str, Smith Hall, Rm 4-105, Newark NJ 07102
WWW:     http://www.linkedin.com/in/yarik        



More information about the Pkg-ExpPsy-PyMVPA mailing list