Reducing RNN Memory Usage by 90%: University of Toronto’s Reversible Neural Networks
Selected from arXiv Authors: Matthew MacKay et al. Translated by: Machine Heart Contributors: Gao Xuan, Zhang Qian Recurrent Neural Networks (RNNs) achieve the best current performance in processing sequential data, but they require a large amount of memory during training. Reversible Recurrent Neural Networks provide a way to reduce the memory requirements for training, as … Read more