question 27

hi,

can someone explain why this is true?
I though Glove was a matrix factorization problem and word2vec used neural nets- how do we know they have the same complexity for one step?

thank you
q27.JPG

Hi :) I am not sure it will help you but: GloVe is indeed a matrix factorisation but you can solve it by SGD or ALS as seen in the notes. For word2vec, it was said during the lecture that you can train it with SGD (you train a binary classifier). But mostly I think the point in this true/false is that for word2vec you train "on the fly", while for GloVe you need to precompute the entire co-occurence matrix which is very big and thus hard do hold in memory. So that is why I would say it is indeed true.

Page 1 of 1

Add comment

Post as Anonymous Dont send out notification