In the example with “california_housing” data set the scoring for learning_curve function was “neg_mean_absolute_error”, whereas in this exercice the scoring keeps its default value i.e. None.
The effect of this parameter is not clear for me. Is the accuracy computed by cv parameter if scoring is None ?
When scoring is not specified, it will default to the R2 score for regression and accuracy for classification.
Here, we deal with a regression problem and we force to use the mean absolute error (the negative mean absolute error indeed).
You can check the following thread to understand the difference between score and error in scikit-learn: Score vs error