Riemannian stochastic optimization methods avoid strict saddle points

Re-assign community