The L2 norm measures a vector’s size by taking the square root of the sum of its squared components. In model regularization, the sum of squared coefficients is used as a measure of model complexity; the model’s bias is left out. section 4.5 section 4.7
Adding this complexity measure to regression error discourages overly complex models. L2 regularization, also called ridge regression, gradually shrinks coefficient values but does not drive them to zero. section 4.5 section 4.7
Ridge regression error = Regression error + L2 norm
Ridge regression combines regression error with the L2 complexity measure. section 4.5
From recommendation models to polynomial regression
A movie recommendation model can be assessed by the sizes of its coefficients, with the bias excluded. An exercise applies the L2 measure to a polynomial regression model and compares its ridge error with lasso error. section 4.5 section 4.8
