This paper proposes a novel approach to domain translation. Leveraging established parallels between generative models and dynamical systems, we propose a reformulation of the Cycle-GAN architecture. By embedding our model with a Hamiltonian structure, we obtain a continuous, expressive and most importantly invertible generative model for domain translation.

翻译：本文提出了一种新颖的域名翻译方法。 利用基因模型和动态系统之间既有的平行关系,我们建议重新制定循环GAN结构。 通过将我们的模型嵌入汉密尔顿结构,我们获得了一种连续的、直观的和最重要的不可忽略的域名翻译基因模型。