Machine Learning Question

uses the whole training set to compute gradients and update the parameters at each however updates the parameters for c) iteration (step) so it is for large data sets.. each training instance. _bounces up and down so it may never settle down at the minimum( but, it gets very close the minimum), while gently marches to the minimum. One solution to settle on the minimum is start with a learning rate and gradually. it. According to the text above, choose the best answer. BGD: Batch Gradient Descent SGD: Stochastic Gradient Descent BGD / slow / SGD / BGD / SGD / small / increase A) B) BGD / slow / BGD / SGD / BGD / small / increase C) BGD / slow / SGD / BGD / SGD / large / reduce D) BGD / slow / SGD/ SGD / BGD / large / reduce E) SGD / fast / BGD / SGD / BGD / small / increase F) SGD/ fast / BGD / BGD / SGD / small / increase G) SGD/ slow / BGD / SGD/ BGD /small / increase H) SGD/ slow / SGD / BGD / SGD / small / increase I) BGD / fast / SGD / SGD / BGD / small / increase J) SGD / fast / BGD / SGD / BGD / large / reduce d) You run gradient descent for 15 iterations with 7=0.3 (learning rate) and compute the cost after each iteration. You find that the cost decreases very slowly during these 15 iterations. Based on this, which of the following conclusions seem most plausible? A) Rather than use the current value of 7, it'd be more promising to try smaller value of 7 (say 7=0.1). B) Rather than use the current value of 7, it'd be more promising to try larger value of 7 (say 7=1.0). C) 7=0.3 is an effective choice of learning rate