Question

Linear Algebra

1. What is stochastic gradient descent in contrast to gradient descent? Why might you choose one versus the other?

Answer 1

based on Descent can one. Since a we datasets, using Gradient Stochastic Gradient Descent (56) în Gradient Descent optimization, we compite the cost gradient the complete training set. In case of very longe be quite costly are only taking single step for over the training set, thus the larger set, the slower algorithm updates the weights and may take untill it converges, to the global cost minimum for more epochs; for each weight j the training pass 00% 1 the longer it . one 07 • wjia wt Bugi, where; sujan Li (target (1) output W')x360) after each training sample Instead, we update the weights for training sample i; for each weight j; I wis wt Awj, where Aw; = 1 (torget (? -outpart?) ; ब Gradient Descent (Go: using the Gradient decent optimization algorithms, the weights updated incrementally after each epoch (= pass over the training dataset), compatible cost function Jo) T6w) - = 5, (torget "2 outputki) ? Augan Jwj (w) Gradient Initial zybian aJ Wi-wtoW Global Y cest minimum Timis (W) sw > vector that contain weights --Wę (target"? output") - 3 - outpat.6)). on & (target (1) ni