On Self Modulation for Generative Adversarial Networks
This addresses the problem of unstable GAN training for researchers and practitioners, offering a simple, incremental improvement that enhances model performance without requiring labeled data or additional tuning.
The paper tackles the challenge of training Generative Adversarial Networks (GANs) by proposing self-modulation, an architectural modification that improves performance across various settings, resulting in a relative decrease of 5%-35% in FID and improved performance in 86% of studied cases.
Training Generative Adversarial Networks (GANs) is notoriously challenging. We propose and study an architectural modification, self-modulation, which improves GAN performance across different data sets, architectures, losses, regularizers, and hyperparameter settings. Intuitively, self-modulation allows the intermediate feature maps of a generator to change as a function of the input noise vector. While reminiscent of other conditioning techniques, it requires no labeled data. In a large-scale empirical study we observe a relative decrease of $5\%-35\%$ in FID. Furthermore, all else being equal, adding this modification to the generator leads to improved performance in $124/144$ ($86\%$) of the studied settings. Self-modulation is a simple architectural change that requires no additional parameter tuning, which suggests that it can be applied readily to any GAN.