Convergence of Alternating Gradient Descent for Matrix Factorization

We consider alternating gradient descent (AGD) with fixed step size applied to the asymmetric matrix factorization objective.