POP-FL: Towards Efficient Federated Learning on Edge Using Parallel Over-Parameterization | IEEE Journals & Magazine | IEEE Xplore