Reversible Recurrent Neural Networks

Abstract

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.32nd Conference on Neural Information Processing Systems (NIPS 2018),

Reversible Vision Transformers

“…Reversible CNNs have been applied to several traditional image tasks such as compression [46], reconstruction [43], retrieval [42], and denoising [33,47] as well as to compressed sensing [61], compact resolution [75], image to image translation [67], remote sensing [56], medical image segmentation [55,74] and MRI reconstruction [57]. Reversible transformation have also been adapted to other networks such as RNNs [51], Unet [4,16], Masked Convolutional Networks [60] and 1000-layer deep Graph Neural Networks [40]. Some early attempts have also been made to adapt the reversible transformation to the NLP domain, initiated by Kiatev et al [38] and built upon in [78,79] for machine translation.…”