What is dropout regularization and what is it used for?
During training, some of the neurons in the neural network are excluded. This technic ensures that the output is not totally dependent on any single neuron. It’s a popular technic to reduce overfitting.