A Dynamic Sliding Window Based Tensor Communication Scheduling Framework for Distributed Deep Learning | IEEE Journals & Magazine | IEEE Xplore