Question 3 2 pts Which of the following is true of stochastic gradient descent but not batch gradient descent? computes the gradient using the entire dataset before weight update O converges much faster because it updates weight more frequently Question 4 3 pts In backprop neural network training the term 'alpha' or momentum refers to: the weight update of the n_th iteration depend partially on the weight update of the (n-1)_th iteration. the effect of ensuring that the learning and weight updates explore all possible directions so as to not miss the global minima. the effect of gradient descent converging too quickly towards a minima