ChainGAN: A sequential approach to GANs

TL;DR

ChainGAN introduces a novel sequential generator architecture for GANs, where initial crude samples are refined through a chain of independent editor networks, enhancing flexibility and efficiency.

Contribution

It presents a new two-step generator architecture with independent training of editors, differing from traditional end-to-end GAN training methods.

Findings

Achieves competitive results on multiple datasets.

Demonstrates robustness and flexibility across architectures.

Improves sample refinement process in GAN training.

Abstract

We propose a new architecture and training methodology for generative adversarial networks. Current approaches attempt to learn the transformation from a noise sample to a generated data sample in one shot. Our proposed generator architecture, called $\textit{ChainGAN}$, uses a two-step process. It first attempts to transform a noise vector into a crude sample, similar to a traditional generator. Next, a chain of networks, called $\textit{editors}$, attempt to sequentially enhance this sample. We train each of these units independently, instead of with end-to-end backpropagation on the entire chain. Our model is robust, efficient, and flexible as we can apply it to various network architectures. We provide rationale for our choices and experimentally evaluate our model, achieving competitive results on several datasets.