Solve Ridge to predict the number of points a Wine will receive. Run Ridge on the training set, with lambda = 0.01, 0.1, 1, 10, 100, 1000. At each solution, record the root-mean-squared-error (RMSE) on training, validation and leave-one-out-cross-validation data.
Plot the train, validation and leave-one-out-cross-validation RMSE values together on a plot against Lambda. Label each curve in the plot.
What lambda achieve the best LOOCV performance? For the model using this lambda, report the objective value, the sum of square errors (on training data), the value of the regularization term.
Using the model that you computed using lambda that achieves best LOOCV performance, list the top 10 most important features and the top 10 least important features. Comment if the weights make sense intuitively.
Use your model to predict the points for the reviews in test data. Report the RMSE.
