Imagine that you have a dataset with continuous values for the dependent variable (eg. the house prices dataset). You want to predict the dependent variable using a single number. This is a constant function predictor - it returns the same constant regardless of the features of the item. What is the best constant function predictor? In this setting, interpret the error, the bias, and the variance. Does it still make sense to use a Train and Test set to learn the constant function predictor?