Accelerating Deep Learning with Shrinkage and Recall | IEEE Conference Publication | IEEE Xplore