Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with
or
.
Download ZIP
Commits on Mar 7, 2011
  1. @yarikoptic
Commits on Mar 4, 2011
  1. @yarikoptic
Commits on Mar 3, 2011
  1. @yarikoptic
  2. @yarikoptic

    BF: sanitize scl_* fields in map2nifti

    yarikoptic authored
    decided to go with minimalistic changes (instead of completely disregarding
    the header but orientations) just to minimize the change from 0.4.6
  3. @yarikoptic
  4. @yarikoptic

    BF+NF: scale data from NIfTI upon load by default (use scale_data=Fal…

    yarikoptic authored
    …se to prevent)
    
    Might be of critical importance for those loading sequences of
    single-volume data files generated by SPM with per-file defined
    scl_parameters.
    
    if loaded from multiple separate volumes, scl_ parameters of the resultant
    Nifti get reset to nominal slope=1.0, inter=0.0
Commits on Mar 2, 2011
  1. @yarikoptic
Commits on Feb 22, 2011
  1. @yarikoptic

    BF: The best error is the lowest ;-)

    Michael W. Cole authored yarikoptic committed
  2. @yarikoptic

    minor - missing space

    yarikoptic authored
Commits on Feb 15, 2011
  1. @yarikoptic

    ENH: unittest to assure that clfs could handle input with a single un…

    yarikoptic authored
    …ique label
    
    some still can't and it is marked as TODO
  2. @yarikoptic
  3. @yarikoptic

    ENH: TreeClassifier could have trailing nodes with no classifier nece…

    yarikoptic authored
    …ssary
    
    e.g. if tree looks like
    
             SVM
             / \
           g1   g2
           /     \
          1     SVM
                /  \
               2    3
Commits on Feb 3, 2011
  1. @yarikoptic
Commits on Feb 1, 2011
  1. @yarikoptic
  2. @yarikoptic

    changelog entry for 0.4.6

    yarikoptic authored
  3. @yarikoptic
  4. @yarikoptic

    ENH: joy of code duplication -- adding the same logic for warnings su…

    yarikoptic authored
    …ppression to mvpa/tests/main.py
  5. @yarikoptic

    enable control over numpy warnings in tests/__init__.py:run -- verbos…

    yarikoptic authored
    …ity>=3 triggers them all on, otherwise - ignore
Commits on Jan 31, 2011
  1. @yarikoptic
  2. @yarikoptic
  3. @yarikoptic

    BF: use centers of the bins as returned by numpy's histogram (not con…

    yarikoptic authored
    …verted only for numpy < 1.1)
  4. @yarikoptic

    Numpy >= 1.4 no longer supports the 'new' kwarg for histogram. Modifi…

    James Kyle authored yarikoptic committed
    …ed test to properly remove the 'new' keyward from transformers.py's call to numpy.histogram.
Commits on Jan 6, 2011
  1. @yarikoptic
  2. @yarikoptic
Commits on Jan 4, 2011
  1. @yarikoptic
Commits on Jan 3, 2011
  1. @yarikoptic

    BF (in maint/0.4): "fixing" example data (mask, bold) up to nibabel w…

    yarikoptic authored
    …arranty of NIfTI standard-compliance
    
    Files were loaded and saved using nibabel and then original dimensionality for
    the mask (2d) was wrong, nibabel dropped z-dimension information, which was
    reincarnated altogether with cal_max, so we have bold and mask exactly the same
    besides dimensionality:
    
    > nifti_tool -mod_hdr -mod_field dim "3 40 20 1 1 1 1 1" -mod_field pixdim "-1.0 3.1 3.75 3.75 2.5 1.0 1.0 1.0" -mod_field srow_z "0.0 0.0 3.75 -0.0" -mod_field cal_max 2623.0 -infiles mask-ed.nii -prefix ed_
    
    $> nifti_tool -diff_hdr -infiles bold.nii.gz mask.nii.gz
      name                offset  nvals  values
      ------------------- ------  -----  ------
      dim                   40      8    4 40 20 1 1452 1 1 1
      dim                   40      8    3 40 20 1 1 1 1 1
    
    as a feature I guess, consecutive load/save in nibabel would prune
    t-dimensionality information
    
    $> nifti_tool -diff_hdr -infiles mask.nii.gz mask_nibabeled.nii.gz
      name                offset  nvals  values
      ------------------- ------  -----  ------
      pixdim                76      8    -1.0 3.1 3.75 3.75 2.5 1.0 1.0 1.0
      pixdim                76      8    -1.0 3.1 3.75 3.75 1.0 1.0 1.0 1.0
  2. @yarikoptic

    externals.exists raiseException argument could have 'always' as the v…

    yarikoptic authored
    …alue
    
    then exception would be thrown despite global configuration.  It is
    necessary whenever given check is used internally within externals for
    dependent externals (e.g. lars dependent on rpy being available), so that
    exception is caught  later on to signal unavailable external
Commits on Dec 26, 2010
  1. @yarikoptic
Commits on Dec 25, 2010
  1. @yarikoptic
  2. @yarikoptic

    ENH: robust MC iterations -- sustain failures of classifiers to be un…

    yarikoptic authored
    …able to train
    
    some permutations might lead to completely degenerate data for
    whenever it comes to a classifier to be trained, e.g. wrapped feature
    selection classifiers -- might come with 0 features selected
    
    now it will simply skip those iterations and spit out warning and
    total number of skipped ones at the end in debug msgs
  3. @yarikoptic
Commits on Dec 8, 2010
  1. @yarikoptic
  2. @yarikoptic

    NF: adopting from 0.5: having base exception class LearnerError

    yarikoptic authored
    also added FailedToPredictError not used atm by 0.4 code base
  3. @yarikoptic
Commits on Nov 1, 2010
  1. @yarikoptic

    ENH/DOC: PLR -- adding to warehouse (for testing), making it more robust

    yarikoptic authored
    * puke FailedToTrainError if degenerate data is provided
    * DOC: ask for reduced of float type
Something went wrong with that request. Please try again.