eelbrain.pipeline.NCRF

class eelbrain.pipeline.NCRF(mu='auto', nlevels=1, n_iter=10, n_iterc=10, n_iterf=100, n_splits=3, tol=0.001, use_ES=False, basis_std=0.0085)[source]

Neuro-Current Response Function estimator

Fits the TRF directly in source space from sensor data using the ncrf package. The data argument of Pipeline.load_trf() must be left unset for NCRF (sensor data is used and localized internally), and the forward solution and noise covariance are loaded automatically.

Parameters:

  • mu (str | float | Sequence[float]) – Regularization parameter ('auto' to determine through cross-validation, or a numeric value / sequence of values).

  • nlevels (int) – Number of levels for the lead-field decomposition.

  • n_iter (int) – Number of iterations.

  • n_iterc (int) – Number of coordinate-descent iterations.

  • n_iterf (int) – Number of FASTA iterations.

  • n_splits (int) – Number of cross-validation splits for mu='auto'.

  • tol (float) – Convergence tolerance.

  • use_ES (bool) – Use the early-stopping strategy.

  • basis_std (float) – Standard deviation of the temporal basis (in seconds).