Improving GANs Using Optimal Transport

2018/03/15 by Salimans, Tim, Zhang, Han, Radford, Alec +1 · 7 citations
#FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML)

paper · doi:10.48550/arxiv.1803.05573

Abstract

We present Optimal Transport GAN (OT-GAN), a variant of generative adversarial nets minimizing a new metric measuring the distance between the generator distribution and the data distribution. This metric, which we call mini-batch energy distance, combines optimal transport in primal form with an energy distance defined in an adversarially learned feature space, resulting in a highly discriminative distance function with unbiased mini-batch gradients. Experimentally we show OT-GAN to be highly stable when trained with large mini-batches, and we present state-of-the-art results on several popular benchmark problems for image generation.

Cited by

• Gradient Flows of Potential Energies in the Geometry of Sinkhorn Divergences
• Sinkhorn-Drifting Generative Models
• Scalable Dynamic Optimal Transport via Distributed Linearized ADMM
• Symmetrized Sinkhorn-Gibbs Inference for Oscillatory Inverse Problems
• A Multi-domain Image Translative Diffusion StyleGAN for Iris Presentation Attack Detection
• Federated Flow Matching
• Generic Event Boundary Detection via Denoising Diffusion

Related

• Discriminator optimal transport
• SAN: Inducing Metrizability of GAN with Discriminative Normalized Linear Layer
• On the Convergence and Robustness of Training GANs with Regularized Optimal Transport
• Minibatch optimal transport distances; analysis and applications
• Image Generation Via Minimizing Fréchet Distance in Discriminator Feature Space