ridge-regression

  • L2 regularization (MSE)
  • Works by adding this term α∗∣coefficients2∣\alpha * |coefficients^2| in the cost function so we are penalizing the size of the coefficients
    • This will uniformly decrease the coefficients
    • The parameter α\alpha will control the shrinkage which makes the coefficients more robust to collinearity
  • The alpha term can be found by using cross-validation techniques
    • sklearn.linear_Regression.RidgeCV uses leave-one-out-cross-validation technique to find the alpha that minimizes the cross-validated mean squared error