Filter Response Normalization Layer: Eliminating Batch Dependence in the Training of Deep Neural Networks
This addresses a key limitation in normalization methods for deep learning, offering a more robust solution for training with varying batch sizes, though it is incremental relative to prior work like Group Normalization.
The paper tackles the problem of batch dependence in deep neural network training by proposing the Filter Response Normalization (FRN) layer, which eliminates dependency on other batch elements and outperforms Batch Normalization and alternatives across various batch sizes, achieving improvements of 0.7-1.0% in ImageNet classification and 0.3-0.5% in COCO object detection.
Batch Normalization (BN) uses mini-batch statistics to normalize the activations during training, introducing dependence between mini-batch elements. This dependency can hurt the performance if the mini-batch size is too small, or if the elements are correlated. Several alternatives, such as Batch Renormalization and Group Normalization (GN), have been proposed to address this issue. However, they either do not match the performance of BN for large batches, or still exhibit degradation in performance for smaller batches, or introduce artificial constraints on the model architecture. In this paper we propose the Filter Response Normalization (FRN) layer, a novel combination of a normalization and an activation function, that can be used as a replacement for other normalizations and activations. Our method operates on each activation channel of each batch element independently, eliminating the dependency on other batch elements. Our method outperforms BN and other alternatives in a variety of settings for all batch sizes. FRN layer performs $\approx 0.7-1.0\%$ better than BN on top-1 validation accuracy with large mini-batch sizes for Imagenet classification using InceptionV3 and ResnetV2-50 architectures. Further, it performs $>1\%$ better than GN on the same problem in the small mini-batch size regime. For object detection problem on COCO dataset, FRN layer outperforms all other methods by at least $0.3-0.5\%$ in all batch size regimes.