Why Regularization?

This article will cover the widely used technique to avoid overfitting. Deep neural networks tend to overfit because of their complexity, large number of hidden layers, where the training error is very small but the testing error is may go up. Regularization helps the model to generalize better so that it performs better with unseen data. Regularization introduces uncertainty or randomness to the learning algorithm, it also simplifies the neural network. Some of the regularization techniques penalize the weight metrics for being too large some techniques reduce the number of hidden units in the Neural Network. There are different types of regularization techniques that affect the model very differently.