Dropout is a simple way to prevent a neural network from overfitting. It is the dropping out of some of the units in a neural network. It is similar to the natural reproduction process, where the nature produces offsprings by combining distinct genes (dropping out others) rather than strengthening the co-adapting of them.