Quasi-Newton Updating for Large-Scale Distributed Learning