R-squared score

The problem, from first principles

Mean squared error tells you how large the errors are, but not whether they are large relative to the spread of the target. A regressor that misses by 5 units is excellent when the target ranges over 1000 and poor when it ranges over 6. The R-squared score fixes that by comparing the model's squared error against the error of the simplest possible model, one that always predicts the mean of the target.

Implement r2_score(y_true, y_pred), which returns 1 - SS_res / SS_tot.

Constraints

• y_true and y_pred must have the same shape, otherwise raise ValueError.
• If y_true is constant, SS_tot is zero and the score is undefined: raise ValueError.
• A perfect prediction scores 1. Predicting the mean scores 0. Worse predictions score below 0.
• Do not modify the inputs.

Hints

Open one at a time. Each gives away a little more than the last.