CHAIN: Enhancing Generalization in Data-Efficient GANs via lipsCHitz continuity constrAIned Normalization

arXiv:2404.00521v1 Announce Type: new
Abstract: Generative Adversarial Networks (GANs) significantly advanced image generation but their performance heavily depends on abundant training data. In scenarios with limited data, GANs often struggle with discriminator overfitting and unstable training. Batch Normalization (BN), despite being known for enhancing generalization and training stability, has rarely been used in the discriminator of Data-Efficient GANs. Our work addresses this gap by identifying a critical flaw in BN: the tendency for gradient explosion during the centering and scaling …

abstract advanced adversarial arxiv continuity cs.cv cs.lg data gans generative generative adversarial networks image image generation networks normalization overfitting performance stability struggle training training data type via