Regularization

Dropout as Implicit Bagging in Deep Learning

A concrete way to understand dropout in deep learning: shared-parameter ensemble training that approximates bagging without training separate models.