CoGANs: Collaborative Generative Adversarial Networks

Omkar Joglekar, Goren Gordon

Keywords: GANs, Multiple generators

TL;DR: We introduce a new method to train multi-generator GANs which manages to beat the state-of-the-art for MNIST

Abstract:

In complex creative scenarios, co-creativity by multiple agents offer great advantages. Each agent has a specific skill set and a set of abilities, which is sometimes not enough to perform a general, large and complex task single-handed.
These kinds of tasks benefit substantially from collaboration. In deep learning applications, data generation is an example of such a complex, potentially multi-modal task. Previous Generative Adversarial Networks (GANs) focused on using a single generator to generate multi-modal datasets, which is sometimes known to face issues such as mode-collapse and failure to converge. The multi-generator based works such as MGAN, MMGAN, MADGAN and AdaGAN either require training a classifier online, the use of complex mixture models or sequentially adding generators, which is computationally complex.
In this work, we present a simple, novel approach of training collaborative GANs (CoGAN), with multiple generators and a single critic/discriminator, without introducing external complexities such as a classifier model. We show that this method of workload division meets the state-of-the-art quality metrics, and makes GAN training robust. We present a proof-of-concept on the MNIST dataset, which has 10 modes of data. The individual generators learn to generate different digits from the distribution, and together learn to generate the whole distribution. We introduce a new component to the generator loss during GAN training, based on the Total Variation Distance (TVD) and show that it significantly improves stability during training and performance over state-of-the-art single generator GANs.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Submission Guidelines: Yes

Please Choose The Closest Area That Your Submission Falls Into: Generative models