Continuous Methods : Hamiltonian Domain Translation

TL;DR

This paper introduces a Hamiltonian-structured, continuous, and invertible generative model for domain translation, reformulating Cycle-GAN to leverage dynamical systems principles for improved expressiveness.

Contribution

It presents a novel Hamiltonian-based reformulation of Cycle-GAN, enabling continuous and invertible domain translation models.

Findings

Hamiltonian structure enhances model invertibility

Continuous modeling improves translation expressiveness

Reformulation outperforms traditional Cycle-GANs

Abstract

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.