What strategies would you use to prevent a deep neural network from overfitting on a relatively small dataset?
