Figure A1.

An outline of how the datasource (DS), feature-preprocessors (FP), and classifier (CL) interact within the cross-validator (CV) object's run_cv_decoding method. The the_cross_validator.run_cv_decoding method works by first generating training and test cross-validation splits using the datasource the_datasource.get_data method (line 4). For each cross-validation split, feature-preprocessing is applied to the data (line 12), and a classifier is trained and tested on this preprocessed data (lines 16–17). The accuracy of the classifier's predictions are assessed (line 20), and this whole procedure can be repeated multiple times (line 1), generating new training and test splits (and potentially also pseudo-populations) on each run. Note: XTr_cv and YTr_cv refer to the training data and labels from all cross-validation splits, and XTe_cv and YTe_cv refer to the test data and labels from all cross-validation splits. Similarly, XTr, YTr, XTe, and YTe, refer to the training and test data and labels from one particular cross-validation split.