Abstract

When using Wasserstein GAN loss function for training generative adversarial networks (GAN), it is theoretically necessary to limit the discriminators’ expressive power (so-called discriminator normalization). Such limitation increases the stability of GAN training at the expense of a less expressive final model. Spectral normalization is one of the normalization algorithms that involves applying a fixed operation independently to each discriminator layer. However, the optimal strength of the discriminator limitation varies for different tasks, which requires a parameterized normalization method. This paper proposes modifications to the spectral normalization algorithm that allow changing the strength of the discriminator limitation. In addition to parameterization, the proposed methods can change the degree of limitation during training, unlike the original algorithm. The quality of the obtained models is explored for each of the proposed methods.

中文翻译：

---

生成模型的自适应谱归一化

摘要

当使用 Wasserstein GAN 损失函数训练生成对抗网络（GAN）时，理论上有必要限制判别器的表达能力（所谓的判别器归一化）。这种限制提高了 GAN 训练的稳定性，但代价是最终模型的表达能力较差。谱归一化是归一化算法之一，涉及对每个鉴别器层独立应用固定操作。然而，对于不同的任务，判别器限制的最佳强度有所不同，这需要参数化的归一化方法。本文提出了对谱归一化算法的修改，允许改变鉴别器限制的强度。除了参数化之外，所提出的方法还可以改变训练期间的限制程度，这与原始算法不同。针对所提出的每种方法，探讨了所获得模型的质量。