Translations:Dropout A Simple Way to Prevent Overfitting/9/en

During training, for each training example and each layer, each neuron's output is independently set to zero with probability $ 1 - p $. If $ h_i $ is the output of neuron $ i $, the dropout operation applies: