Sparsity Aware Normalization for GANs
This work addresses the stabilization of GAN training, particularly for image-to-image translation, offering an incremental improvement over existing normalization methods.
The paper tackled the problem of stabilizing GAN training by identifying a drawback in spectral normalization and introduced sparsity aware normalization (SAN), which accounts for sparse feature maps in convolutional networks with ReLU activations. The result showed that SAN improves upon existing methods in image-to-image translation settings, achieving better performance with less training epochs and smaller networks while requiring minimal computational overhead.
Generative adversarial networks (GANs) are known to benefit from regularization or normalization of their critic (discriminator) network during training. In this paper, we analyze the popular spectral normalization scheme, find a significant drawback and introduce sparsity aware normalization (SAN), a new alternative approach for stabilizing GAN training. As opposed to other normalization methods, our approach explicitly accounts for the sparse nature of the feature maps in convolutional networks with ReLU activations. We illustrate the effectiveness of our method through extensive experiments with a variety of network architectures. As we show, sparsity is particularly dominant in critics used for image-to-image translation settings. In these cases our approach improves upon existing methods, in less training epochs and with smaller capacity networks, while requiring practically no computational overhead.