Previous | Next --- Slide 63 of 67
Back to Lecture Thumbnails

Since we are just setting step size to be small without step control, does this mean we may need a lot of iterations? Otherwise we may not reach the minimum.


There are many second-order optimizers that can speed up the optimization, comparing to the first-order methods like gradient decent, one is ADAM which is often used in DNN.

Please log in to leave a comment.