Guiding the One-to-one Mapping in CycleGAN via Optimal Transport
This work addresses a fundamental limitation in CycleGAN for researchers and practitioners in unsupervised learning, offering a more controllable approach, though it is incremental as it builds on existing methods.
The paper tackles the lack of theoretical guarantees for the one-to-one mapping learned by CycleGAN in unsupervised data translation, finding it can be random, and proposes using optimal transport with task-specific constraints to control and improve this mapping, with experiments showing it learns mappings with desired properties.
CycleGAN is capable of learning a one-to-one mapping between two data distributions without paired examples, achieving the task of unsupervised data translation. However, there is no theoretical guarantee on the property of the learned one-to-one mapping in CycleGAN. In this paper, we experimentally find that, under some circumstances, the one-to-one mapping learned by CycleGAN is just a random one within the large feasible solution space. Based on this observation, we explore to add extra constraints such that the one-to-one mapping is controllable and satisfies more properties related to specific tasks. We propose to solve an optimal transport mapping restrained by a task-specific cost function that reflects the desired properties, and use the barycenters of optimal transport mapping to serve as references for CycleGAN. Our experiments indicate that the proposed algorithm is capable of learning a one-to-one mapping with the desired properties.