interfaces.afni.svm

SVMTest

Link to code

Wraps command 3dsvm

Temporally predictive modeling with the support vector machine SVM Test Only For complete details, see the 3dsvm Documentation.

Examples

>>> from nipype.interfaces import afni as afni
>>> svmTest = afni.SVMTest()
>>> svmTest.inputs.in_file= 'run2+orig'
>>> svmTest.inputs.model= 'run1+orig_model'
>>> svmTest.inputs.testlabels= 'run2_categories.1D'
>>> svmTest.inputs.out_file= 'pred2_model1'
>>> res = svmTest.run() 

Inputs:

[Mandatory]
in_file: (an existing file name)
        A 3D or 3D+t AFNI brik dataset to be used for testing.
        flag: -testvol %s
model: (a unicode string)
        modname is the basename for the brik containing the SVM model
        flag: -model %s

[Optional]
args: (a unicode string)
        Additional parameters to the command
        flag: %s
classout: (a boolean)
        Flag to specify that pname files should be integer-valued,
        corresponding to class category decisions.
        flag: -classout
environ: (a dictionary with keys which are a newbytes or None or a
         newstr or None and with values which are a newbytes or None or a
         newstr or None, nipype default value: {})
        Environment variables
ignore_exception: (a boolean, nipype default value: False)
        Print an error message instead of throwing an exception in case the
        interface fails to run
multiclass: (a boolean)
        Specifies multiclass algorithm for classification
        flag: -multiclass %s
nodetrend: (a boolean)
        Flag to specify that pname files should not be linearly detrended
        flag: -nodetrend
nopredcensord: (a boolean)
        Flag to prevent writing predicted values for censored time-points
        flag: -nopredcensord
options: (a unicode string)
        additional options for SVM-light
        flag: %s
out_file: (a file name)
        filename for .1D prediction file(s).
        flag: -predictions %s
outputtype: (u'NIFTI_GZ' or u'AFNI' or u'NIFTI')
        AFNI output filetype
terminal_output: (u'stream' or u'allatonce' or u'file' or u'none')
        Control terminal output: `stream` - displays to terminal immediately
        (default), `allatonce` - waits till command is finished to display
        output, `file` - writes output to file, `none` - output is ignored
testlabels: (an existing file name)
        *true* class category .1D labels for the test dataset. It is used to
        calculate the prediction accuracy performance
        flag: -testlabels %s

Outputs:

out_file: (an existing file name)
        output file

SVMTrain

Link to code

Wraps command 3dsvm

Temporally predictive modeling with the support vector machine SVM Train Only For complete details, see the 3dsvm Documentation.

Examples

>>> from nipype.interfaces import afni as afni
>>> svmTrain = afni.SVMTrain()
>>> svmTrain.inputs.in_file = 'run1+orig'
>>> svmTrain.inputs.trainlabels = 'run1_categories.1D'
>>> svmTrain.inputs.ttype = 'regression'
>>> svmTrain.inputs.mask = 'mask.nii'
>>> svmTrain.inputs.model = 'model_run1'
>>> svmTrain.inputs.alphas = 'alphas_run1'
>>> res = svmTrain.run() 

Inputs:

[Mandatory]
in_file: (an existing file name)
        A 3D+t AFNI brik dataset to be used for training.
        flag: -trainvol %s
ttype: (a unicode string)
        tname: classification or regression
        flag: -type %s

[Optional]
alphas: (a file name)
        output alphas file name
        flag: -alpha %s
args: (a unicode string)
        Additional parameters to the command
        flag: %s
censor: (an existing file name)
        .1D censor file that allows the user to ignore certain samples in
        the training data.
        flag: -censor %s
environ: (a dictionary with keys which are a newbytes or None or a
         newstr or None and with values which are a newbytes or None or a
         newstr or None, nipype default value: {})
        Environment variables
ignore_exception: (a boolean, nipype default value: False)
        Print an error message instead of throwing an exception in case the
        interface fails to run
kernel: (a unicode string)
        string specifying type of kernel function:linear, polynomial, rbf,
        sigmoid
        flag: -kernel %s
mask: (an existing file name)
        byte-format brik file used to mask voxels in the analysis
        flag: -mask %s, position: -1
max_iterations: (an integer (int or long))
        Specify the maximum number of iterations for the optimization.
        flag: -max_iterations %d
model: (a file name)
        basename for the brik containing the SVM model
        flag: -model %s
nomodelmask: (a boolean)
        Flag to enable the omission of a mask file
        flag: -nomodelmask
options: (a unicode string)
        additional options for SVM-light
        flag: %s
out_file: (a file name)
        output sum of weighted linear support vectors file name
        flag: -bucket %s
outputtype: (u'NIFTI_GZ' or u'AFNI' or u'NIFTI')
        AFNI output filetype
terminal_output: (u'stream' or u'allatonce' or u'file' or u'none')
        Control terminal output: `stream` - displays to terminal immediately
        (default), `allatonce` - waits till command is finished to display
        output, `file` - writes output to file, `none` - output is ignored
trainlabels: (an existing file name)
        .1D labels corresponding to the stimulus paradigm for the training
        data.
        flag: -trainlabels %s
w_out: (a boolean)
        output sum of weighted linear support vectors
        flag: -wout

Outputs:

alphas: (a file name)
        output alphas file name
model: (a file name)
        brik containing the SVM model file name
out_file: (a file name)
        sum of weighted linear support vectors file name