Connect your moderator Slack workspace to receive post notifications:
Sign in with Slack

lecture13a_matrix_factorization

Hello,
Is it possible to have the details of the computation of the optimal expressions of Z^T and W^T, obtained by computing the gradient of the loss wrt to Z and W respectively ( from "Alternative Least-squares" part from lecture 13_a).

screenshot-2.jpg

Thank you in advance.

Top comment

You compute the gradient with respect to X and Z then set them to zero, after transposing you will get the expressions above.
All of this can be translated into the ALS algorithm which fixes one matrix and optimizes over the other one. I gave a detailed answer of how to get the update of X in an other question that was asked in this forum, I invite you to refer to it.

Page 1 of 1

Add comment

Post as Anonymous Dont send out notification