ALS and SGD for matrix factorization (MCQ17)

Hi! I am not sure to have understood what is exactly the computational complexity of ALS method, so I can't understand the comparison with SGD. Moreover, why SGD in this case is independent of D and N? Isn't its computational cost in the order of O(D)?

Same question here !

according to me the complexity of als is O(nb_non_zeros*k ) as it is finally the sum stoch gradient nb_non_zeros.

I don't understand why the question asks about the computational complexity of ALS per iteration, if we have closed-form solution for ALS, not iterative? That is, the solution will be obtained in one step.

Computational cost is for SGD is O(K)

Page 1 of 1

Add comment

Post as Anonymous Dont send out notification