High-Performance Large-Scale Image Recognition Without Normalization

             Andrew Brock 1 Soham De 1 Samuel L. Smith 1 Karen Simonyan 1

             Abstract
    Batch normalization is a key component of most
    image classification models, but it has many
    undesirable properties stemming from its depen-
    dence on the batch size and interactions between
    examples. Although recent work has succeeded
    in training deep ResNets without normalization
    layers, these models do  ...