What regularization term does ridge regression use?

Ridge regression incorporates the ℓ₂ norm as its regularization term, which serves to penalize the sum of the squares of the coefficients associated with the features in the regression model. This approach effectively discourages large weights by adding a penalty proportional to the square of the magnitude of coefficients. This results in a more stable and generalized model, particularly in situations where multicollinearity is present among the independent variables or when the number of predictors exceeds the number of observations.

By utilizing the ℓ₂ norm, ridge regression helps prevent overfitting, allowing the model to maintain better predictive performance on unseen data. The penalty term is directly integrated into the loss function, combined with the standard loss (which could be the mean squared error), leading to an overall minimized error across both the prediction accuracy and complexity of the model.

The other choices represent different concepts not aligned with the regularization method used in ridge regression. Mean absolute error and mean squared error are both loss functions rather than regularization terms, while the ℓ₁ norm corresponds to the regularization used in lasso regression, not ridge regression.