Efficient approaches for escaping higher order saddle points in non-convex optimization

Open in new window