Dropout

Dropout is a technique used in training neural networks that helps prevent overfitting. During training, dropout randomly ignores, or 'drops out,' a subset of neurons in a layer, which helps the network to generalize better to new data.