주요 내용1. 개념- randomly drop units during training(temporarily removing it from the network) * explore different regions of the weight space- each unit is retained with p, chosen using a validation set- the individual models are different from each other - At test time, the weights are multiplied by p- The gradients for each parameter are averaged over the training cases in each mini-batch 2. 활용-..
