Which strategy does not prevent a model from over$-$fitting to the training data?

Group of answer choices

Early Stopping

Drop out

Data Augmentation

Pooling