ooc_cuDNN: Accommodating convolutional neural networks over GPU memory capacity | IEEE Conference Publication | IEEE Xplore