Dropout

Dropout is a regularization technique used to prevent overfitting. During each iteration of the training process, some neurons are randomly selected with a given probability (Dropout Rate P) for each layer. The selected neurons are ignored (dropped-out) by setting their activations to zero