Live Gradient Compensation for Evading Stragglers in Distributed Learning | IEEE Conference Publication | IEEE Xplore