Determine an easily computed upper bound on the Lipschitz constant of the gradient of this cost function. Your final expression must not involve any singular values, because the SVD is too expensive to compute for largescale problems. Your bound should be one that is reasonably tight (see below); if your bound is too loose (too large), then using its reciprocal as the step size in gradient descent (or related algorithms) would lead to undesirably slow convergence