skbold.pipelines package

The pipelines module contains some standard MVPA pipelines using the scikit-learn style Pipeline objects.

create_ftest_kbest_svm(kernel='linear', k=100, **kwargs)

Creates an svm-pipeline with f-test feature selection.

Uses SelectKBest from scikit-learn.feature_selection.

Parameters:

• kernel (str) – Kernel for SVM (default: ‘linear’)
• k (int) – How many voxels to select (from the k best)
• **kwargs – Arbitrary keyword arguments for SVC() initialization.

Returns:

ftest_svm – Pipeline with f-test feature selection and svm.

Return type:

scikit-learn Pipeline object

create_ftest_percentile_svm(kernel='linear', perc=10, **kwargs)

Creates an svm-pipeline with f-test feature selection.

Uses SelectPercentile from scikit-learn.feature_selection.

Parameters:

• kernel (str) – Kernel for SVM (default: ‘linear’)
• perc (int or float) – Percentage of voxels to select
• **kwargs – Arbitrary keyword arguments for SVC() initialization.

Returns:

ftest_svm – Pipeline with f-test feature selection and svm.

Return type:

scikit-learn Pipeline object

create_pca_svm(kernel='linear', n_comp=10, whiten=False, **kwargs)

Creates an svm-pipeline with f-test feature selection.

Parameters:

• kernel (str) – Kernel for SVM (default: ‘linear’)
• n_comp (int) – How many PCA-components to select
• whiten (bool) – Whether to use whitening in PCA
• **kwargs – Arbitrary keyword arguments for SVC() initialization.

Returns:

pca_svm – Pipeline with PCA feature extraction and svm.

Return type:

scikit-learn Pipeline object