We usually choose a mini-batch size greater than 1 and less than mm, because that way we make use of vectorization but not fall into the slower case of batch gradient descent.true/false