Reversible Recurrent Neural Networks
This addresses memory limitations in training RNNs for sequential data processing, offering a practical improvement for researchers and practitioners, though it is incremental as it builds on reversible neural network concepts.
The paper tackles the memory-intensive training of recurrent neural networks (RNNs) by introducing reversible RNNs that reduce activation memory costs by factors of 10-15 while maintaining comparable performance, and extends this to attention-based models with similar reductions.
Recurrent neural networks (RNNs) provide state-of-the-art performance in processing sequential data but are memory intensive to train, limiting the flexibility of RNN models which can be trained. Reversible RNNs---RNNs for which the hidden-to-hidden transition can be reversed---offer a path to reduce the memory requirements of training, as hidden states need not be stored and instead can be recomputed during backpropagation. We first show that perfectly reversible RNNs, which require no storage of the hidden activations, are fundamentally limited because they cannot forget information from their hidden state. We then provide a scheme for storing a small number of bits in order to allow perfect reversal with forgetting. Our method achieves comparable performance to traditional models while reducing the activation memory cost by a factor of 10--15. We extend our technique to attention-based sequence-to-sequence models, where it maintains performance while reducing activation memory cost by a factor of 5--10 in the encoder, and a factor of 10--15 in the decoder.