Uncertainty masks

class alpaca.ue.masks.BaseMask

The base class for masks

copy() → alpaca.ue.masks.BaseMask

Creates the copy of an instance

class alpaca.ue.masks.BasicBernoulliMask

The implementation of Monte Carlo Dropout (MCD) logic

More about the behaviours of MCD can be found in: https://arxiv.org/pdf/2008.02627.pdf

Examples

>>> estimator = MCDUE(model, nn_runs=100, acquisition='std')
>>> estimations1 = estimator.estimate(x_batch)

class alpaca.ue.masks.DPPMask(ht_norm: bool = False, covariance: bool = False)

class alpaca.ue.masks.DecorrelationMask(*, scaling: bool = False, ht_norm: bool = False, eps: float = 1e-08)

TODO:

class alpaca.ue.masks.DecorrelationMaskScaled(*args, **kwargs)

TODO:

class alpaca.ue.masks.LeverageScoreMask(*, ht_norm: bool = True, lambda_: int = 1, covariance: bool = False)

TODO:

class alpaca.ue.masks.LeverageScoreMaskCov(*args, **kwargs)

class alpaca.ue.masks.MaskLayered(*args, **kwargs)

The base class for nn layered masks