DropoutUncertainty#

class pinnx.callbacks.DropoutUncertainty(period=1000)[source]#

Uncertainty estimation via MC dropout.

References

Y. Gal, & Z. Ghahramani. Dropout as a Bayesian approximation: Representing trainer uncertainty in deep learning. International Conference on Machine Learning, 2016.

Warning

This cannot be used together with other techniques that have different behaviors during training and testing, such as batch normalization.

on_epoch_end()[source]#

Called at the end of every epoch.

on_train_end()[source]#

Called at the end of trainer training.