Normalizing the Normalizers: Comparing and Extending Network Normalization Schemes
This work addresses the problem of inconsistent normalization performance across network architectures for researchers and practitioners in deep learning, representing an incremental but practical extension.
The paper proposes a unified divisive normalization framework that encompasses batch and layer normalization, and shows that adding a sparse regularizer to these normalization schemes yields significant improvements over standard techniques in convolutional and recurrent neural networks.
Normalization techniques have only recently begun to be exploited in supervised learning tasks. Batch normalization exploits mini-batch statistics to normalize the activations. This was shown to speed up training and result in better models. However its success has been very limited when dealing with recurrent neural networks. On the other hand, layer normalization normalizes the activations across all activities within a layer. This was shown to work well in the recurrent setting. In this paper we propose a unified view of normalization techniques, as forms of divisive normalization, which includes layer and batch normalization as special cases. Our second contribution is the finding that a small modification to these normalization schemes, in conjunction with a sparse regularizer on the activations, leads to significant benefits over standard normalization techniques. We demonstrate the effectiveness of our unified divisive normalization framework in the context of convolutional neural nets and recurrent neural networks, showing improvements over baselines in image classification, language modeling as well as super-resolution.