[pymvpa] ENET error

Thu Sep 29 15:15:34 UTC 2016

Thanks!
I now use the remove_invariant_features() on my full dataset first, and then run the program again.
But I still got the same error message:

/lustre/work/apps/anaconda/lib/python2.7/site-packages/mvpa2/measures/corrstability.py:94: RuntimeWarning: invalid value encountered in divide
  covar = (dat1*dat2).mean(0) / (dat1.std(0) * dat2.std(0))
/lustre/work/apps/anaconda/lib/python2.7/site-packages/rpy2/rinterface/__init__.py:185: RRuntimeWarning: Error in y - mu : non-numeric argument to binary operator

And warning message:
WARNING: Failed to obtain value from <CrossValidation> due to Failed to predict on <Dataset: 4956x5000 at float64, <sa: chunks,partitions,targets,time_coords,time_indices>, <fa: voxel_indices>, <a: imgaffine,imghdr,imgtype,lastpartitionset,lastsplit,mapper,partitions_set,repetitons...> using <ENET>. Exceptions was: Error in y - mu : non-numeric argument to binary operator
.  Measurement was skipped, which could lead to unstable and/or incorrect assessment of the null_dist

I did not use any mask this time.
Any thoughts on that? Any help will be appreciated!

From: Yaroslav Halchenko [mailto:yoh at onerussian.com]
Sent: Tuesday, September 20, 2016 4:43 PM
To: Development and support of PyMVPA <pkg-exppsy-pymvpa at lists.alioth.debian.org>; Liang, Guangsheng <guangsheng.liang at ttu.edu>
Subject: Re: [pymvpa] ENET error

On September 20, 2016 3:46:40 PM EDT, "Liang, Guangsheng" <guangsheng.liang at ttu.edu<mailto:guangsheng.liang at ttu.edu>> wrote:
Hello,

Thank you very much for your kindly response.
Which function that I can remove them?

Thanks again!
Carl

From: Pkg-ExpPsy-PyMVPA [mailto:pkg-exppsy-pymvpa-bounces+guangsheng.liang=ttu.edu at lists.alioth.debian.org] On Behalf Of Richard Dinga
Sent: Tuesday, September 20, 2016 3:50 AM
To: Development and support of PyMVPA <pkg-exppsy-pymvpa at lists.alioth.debian.org<mailto:pkg-exppsy-pymvpa at lists.alioth.debian.org>>
Subject: Re: [pymvpa] ENET error

I guess you have invariant features in your dataset, therefore you will get problems when trying to divide by 0. There is a function to remove them.

On Fri, Sep 16, 2016 at 8:01 PM, Liang, Guangsheng <guangsheng.liang at ttu.edu<mailto:guangsheng.liang at ttu.edu>> wrote:
Hello PyMVPA community,

I am currently working on a fMRI data applying an ENET classifier.
My PyMVPA is running under Linux environment, with python 2.7.12.
The design of the data is a repeated measurement with pre and post conditions.
In my understanding, in my case, chunk value should store subject ID, target value should store the time conditions.
I am using the classifier code that I found in the maillist: http://lists.alioth.debian.org/pipermail/pkg-exppsy-pymvpa/2009q1/000412.html

clf = FeatureSelectionClassifier(\
                ENET(lm=1.0,max_steps=500,trace=False,normalize=False),\
                SensitivityBasedFeatureSelection(\
                                CorrStability(),\
                                FixedNElementTailSelector(5000,mode='select',tail='upper')),\
                descr="ENET on 5K best(CorrStability) features")

I am also using 1000 times of permutation to test the null hypothesis, which are almost the same as those in the tutorial: http://www.pymvpa.org/tutorial_significance.html#the-following-content-is-incomplete-and-experimental (section: Avoiding the trap OR Advanced magic 101)

However, errors suspend my program:

/lustre/work/apps/anaconda/lib/python2.7/site-packages/mvpa2/measures/corrstability.py:94: RuntimeWarning: invalid value encountered in divide
  covar = (dat1*dat2).mean(0) / (dat1.std(0) * dat2.std(0))
/lustre/work/apps/anaconda/lib/python2.7/site-packages/rpy2/rinterface/__init__.py:185: RRuntimeWarning: Error in y - mu : non-numeric argument to binary operator

  warnings.warn(x, RRuntimeWarning)
Traceback (most recent call last):
  File "15subj_enet.py", line 68, in <module>
    err_fds = cv_mc_corr_fds(fds)
  File "/lustre/work/apps/anaconda/lib/python2.7/site-packages/mvpa2/base/learner.py", line 258, in __call__
    return super(Learner, self).__call__(ds)
  File "/lustre/work/apps/anaconda/lib/python2.7/site-packages/mvpa2/base/node.py", line 136, in __call__
    self._precall(ds)
  File "/lustre/work/apps/anaconda/lib/python2.7/site-packages/mvpa2/measures/base.py", line 120, in _precall
    self.__null_dist.fit(measure, ds)
  File "/lustre/work/apps/anaconda/lib/python2.7/site-packages/mvpa2/clfs/stats.py", line 427, in fit
    % (measure, skipped))
RuntimeError: Failed to obtain any value from <CrossValidation>. 1000 measurements were skipped. Check above warnings, and your code/data

I was wondering if someone could kindly explain what are those mean, and how do I resolve this?

Thank you very much!
Carl

_______________________________________________
Pkg-ExpPsy-PyMVPA mailing list
Pkg-ExpPsy-PyMVPA at lists.alioth.debian.org<mailto:Pkg-ExpPsy-PyMVPA at lists.alioth.debian.org>
http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-exppsy-pymvpa

________________________________

Pkg-ExpPsy-PyMVPA mailing list
Pkg-ExpPsy-PyMVPA at lists.alioth.debian.org<mailto:Pkg-ExpPsy-PyMVPA at lists.alioth.debian.org>
http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-exppsy-pymvpa

http://www.pymvpa.org/faq.html#is-there-an-easy-way-to-remove-invariant-features-from-a-dataset
;-)
--
Sent from a phone which beats iPhone.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.alioth.debian.org/pipermail/pkg-exppsy-pymvpa/attachments/20160929/11ffd5e7/attachment-0001.html>