ncrf.crossvalidate

ncrf.crossvalidate(data, candidates, cv)

Perform cross-validation over a set of solver candidates.

Each candidate is fit and scored on the same folds, and the resulting CVResult objects are returned for the caller to compare. This is what search() calls to score the configurations it chooses between.

Parameters:
  • data – Whitened M/EEG data and the corresponding stimulus variables (see RegressionData.whiten()); its forward model is the one the candidates are fit with. It is pickled to send it to worker processes.

  • candidates – Fixed solver configurations to compare.

  • cv – Folds and worker count to score the candidates on.

Returns:

Cross-validation results.

Return type:

list