Updates parameters after computing the gradient using the entire training dataset in each iteration.

Updates parameters without computing gradients.

Updates parameters after computing the gradient on a random subset (mini-batch) of training data.

Updates parameters after computing the gradient on a single training example.