Improving neural networks by preventing co-adaptation of feature detectors

Geoffrey E. Hinton, Nitish Srivastava, Alex Krizhevsky, Ilya Sutskever, and Ruslan R. Salakhutdinov submitted the paper 'Improving neural networks by preventing co-adaptation of feature detectors' on July 3, 2012. The paper describes reducing overfitting by randomly omitting feature detectors during training and refers to the method as random dropout.