deephyper.ensemble.loss

deephyper.ensemble.loss#

Subpackage for loss functions of ensemble models.

Classes

AbsoluteError

The usual absolute loss (y_true - y_pred)**2 used to estimate the median of P(Y|X=x).

CategoricalCrossEntropy

Categorical-Cross Entropy (a.k.a., Log-Loss) function for classification.

Loss

Base class that represents the loss function of an ensemble.

NormalNegLogLikelihood

The negative log-likelihood of observed data y_true given predicted parameters of a normal distribution y_pred["loc"], y_pred["scale"].

SquaredError

The usual square loss (y_true - y_pred)**2 used to estimate E[Y|X=x].

ZeroOneLoss

Zero-One loss for classification (a.k.a, error rate) which is 1 if the prediction is wrong and 0 if it is correct.