Regularized least squares
	The penalty on the squared weights is mathematically compatible with the squared error function, so we get a nice closed form for the optimal weights with this regularizer:
				identity matrix