Regularization is a technique for reducing the variance in the validation set, thus preventing the model from overfitting during training. In doing so, the model can better generalize to new examples. When training deep neural networks, a couple of strategies exist for use as a regularizer.
Dropout
Dropout is a regularization technique that prevents a deep neural network from overfitting by randomly discarding a number of neurons at every layer during training. In doing so, the neural network is not overly dominated by any one ...
