WeightAlign: Normalizing Activations by Weight Alignment
This addresses a critical bottleneck in training deep neural networks with small batches, offering a stable alternative to existing normalization methods.
The paper tackles the instability of batch normalization at small batch sizes by introducing WeightAlign, a method that normalizes activations using weight statistics instead of sample statistics, achieving stable performance across various batch sizes and tasks including classification, segmentation, and domain adaptation.
Batch normalization (BN) allows training very deep networks by normalizing activations by mini-batch sample statistics which renders BN unstable for small batch sizes. Current small-batch solutions such as Instance Norm, Layer Norm, and Group Norm use channel statistics which can be computed even for a single sample. Such methods are less stable than BN as they critically depend on the statistics of a single input sample. To address this problem, we propose a normalization of activation without sample statistics. We present WeightAlign: a method that normalizes the weights by the mean and scaled standard derivation computed within a filter, which normalizes activations without computing any sample statistics. Our proposed method is independent of batch size and stable over a wide range of batch sizes. Because weight statistics are orthogonal to sample statistics, we can directly combine WeightAlign with any method for activation normalization. We experimentally demonstrate these benefits for classification on CIFAR-10, CIFAR-100, ImageNet, for semantic segmentation on PASCAL VOC 2012 and for domain adaptation on Office-31.