The Reversible Residual Network: Backpropagation Without Storing Activations

Abstract

Deep residual networks (ResNets) have significantly pushed forward the state-of-the-art on image classification, increasing in performance as networks grow both deeper and wider. However, memory consumption becomes a bottleneck, as one needs to store the activations in order to calculate gradients using backpropagation. We present the Reversible Residual Network (RevNet), a variant of ResNets where each layer's activations can be reconstructed exactly from the next layer's. Therefore, the activations for most layers need not be stored in memory during backpropagation. We demonstrate the effectiveness of RevNets on CIFAR-10, CIFAR-100, and ImageNet, establishing nearly identical classification accuracy to equally-sized ResNets, even though the activation storage requirements are independent of depth.

References in corpus (5)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Revisiting Distributed Synchronous SGD
• Gradient-based Hyperparameter Optimization through Reversible Learning
• Neural Machine Translation in Linear Time
• Wider or Deeper: Revisiting the ResNet Model for Visual Recognition

Cited by in corpus (6)

• Optimal checkpointing for heterogeneous chains: how to train deep neural networks with limited memory
• Neural Invertible Variable-degree Optical Aberrations Correction
• Training Invertible Neural Networks as Autoencoders
• Initialization Matters: Regularizing Manifold-informed Initialization for Neural Recommendation Systems
• Memory-efficient Learning for High-Dimensional MRI Reconstruction
• Depthwise Separable Convolutions Allow for Fast and Memory-Efficient Spectral Normalization