Keynote Lecture : Gradient compression for efficient distributed deep learning | IEEE Conference Publication | IEEE Xplore