Regularization is the technique of adding a parameter, λ, to the loss function of a learning algorithm to improve its ability to generalize to new examples by reducing overfitting. The role of the extra regularization parameter is to shrink or to minimize the measure of the weights (or parameters) of the other features in the model.
Regularization is applied to linear models such as polynomial linear regression and logistic regression which are susceptible to overfitting when high-order polynomial features are added to the set ...
