Exploring Weight Symmetry in Deep Neural Networks
arXiv:1812.11027
Abstract
We propose to impose symmetry in neural network parameters to improve parameter usage and make use of dedicated convolution and matrix multiplication routines. Due to significant reduction in the number of parameters as a result of the symmetry constraints, one would expect a dramatic drop in accuracy. Surprisingly, we show that this is not the case, and, depending on network size, symmetry can have little or no negative effect on network accuracy, especially in deep overparameterized networks. We propose several ways to impose local symmetry in recurrent and convolutional neural networks, and show that our symmetry parameterizations satisfy universal approximation property for single hidden layer networks. We extensively evaluate these parameterizations on CIFAR, ImageNet and language modeling datasets, showing significant benefits from the use of symmetry. For instance, our ResNet-101 with channel-wise symmetry has almost 25% less parameters and only 0.2% accuracy loss on ImageNet. Code for our experiments is available at https://github.com/hushell/deep-symmetry
References in corpus (12)
- MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications
- Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour
- Recurrent Neural Network Regularization
- Densely Connected Convolutional Networks
- Exploiting Linear Structure Within Convolutional Networks for Efficient Evaluation
- Training Very Deep Networks
- Character-Aware Neural Language Models
- Aggregated Residual Transformations for Deep Neural Networks
- Highway and Residual Networks learn Unrolled Iterative Estimation
- Soft Weight-Sharing for Neural Network Compression
- ShaResNet: reducing residual network parameter number by sharing weights
- Necessary And Sufficient Conditions For Existence of the LU Factorization of an Arbitrary Matrix