eelbrain.pipeline.Boosting

class eelbrain.pipeline.Boosting(basis=0.05, basis_window='hamming', error='l1', delta=0.005, mindelta=None, selective_stopping=0, scale_data='inplace', partitions=None, cv=True, partition_results=False, backward=False)[source]

Boosting estimator

Parameters:

  • tstart – Not set here; see Pipeline.load_trf().

  • basis (float) – Width of the basis window for the response function in seconds.

  • basis_window (str) – Window shape for the basis (see eelbrain.boosting()).

  • error (Literal['l1', 'l2']) – Error function: 'l1' or 'l2'.

  • delta (float) – Boosting step size.

  • mindelta (float) – If the error for the training data can’t be reduced, divide delta in half until it is smaller than mindelta.

  • selective_stopping (int) – Stop boosting each predictor separately (see eelbrain.boosting()).

  • scale_data (bool | Literal['inplace']) – Scale y and x before fitting; 'inplace' to save memory.

  • partitions (int) – Number of partitions for cross-validation. None to infer from the number of cases; a negative value concatenates the cases and uses -partitions partitions (-1 to let boosting infer them).

  • cv (bool) – Use cross-validation (hold out a test partition).

  • partition_results (bool) – Keep the result for each test partition.

  • backward (bool) – Fit a backward model (predict the stimulus from the response). Only valid with a single-term model.