ott.neural.methods.monge_gap.MongeGapEstimator.regularizer

property MongeGapEstimator.regularizer: Callable[[Array, Array], float]

Regularizer added to the fitting loss.

Can be, e.g. the monge_gap_from_samples(). If no regularizer is passed for solver instantiation, or regularization weight regularizer_strength is 0, return 0 by default along with an empty set of log values.