Paired Wasserstein Autoencoders for Conditional Sampling
This addresses a theoretical bottleneck for researchers in generative modeling, offering a method for conditional sampling in imaging tasks, though it appears incremental as it builds on existing Wasserstein autoencoder frameworks.
The paper tackled the theoretical difficulties of adapting Wasserstein autoencoders to conditional sampling by proposing paired autoencoders, enabling tasks like denoising, inpainting, and unsupervised image translation with practical applicability demonstrated in experiments.
Wasserstein distances greatly influenced and coined various types of generative neural network models. Wasserstein autoencoders are particularly notable for their mathematical simplicity and straight-forward implementation. However, their adaptation to the conditional case displays theoretical difficulties. As a remedy, we propose the use of two paired autoencoders. Under the assumption of an optimal autoencoder pair, we leverage the pairwise independence condition of our prescribed Gaussian latent distribution to overcome this theoretical hurdle. We conduct several experiments to showcase the practical applicability of the resulting paired Wasserstein autoencoders. Here, we consider imaging tasks and enable conditional sampling for denoising, inpainting, and unsupervised image translation. Moreover, we connect our image translation model to the Monge map behind Wasserstein-2 distances.