deephyper.ensemble

Ensemble subpackage.

This subpackage provides tools to build ensemble of predictive models. Ensembling can be useful to minimize the loss if the learner has variance (i.e., its performance metric varies when trained multiple times) or to estimate the uncertainty of the predictions.

Classes

EnsemblePredictor

A predictor that is itself an ensemble of multiple predictors.

aggregator	Subpackage with aggregation functions for the predictions of set of predicting models.
loss	Subpackage for loss functions of ensemble models.
selector	Subpackage providing ensemble selection algorithms.