Latent reweighting, an almost free improvement for GANs
This work addresses the issue of image quality degradation in GANs for machine learning practitioners, offering an incremental improvement by enhancing sampling efficiency.
The paper tackles the problem of low-quality image sampling in GANs due to misspecified latent spaces by introducing latent importance weights and sampling methods to avoid poor samples, resulting in improved sampling quality without increased computational cost, as demonstrated on synthetic and high-dimensional datasets.
Standard formulations of GANs, where a continuous function deforms a connected latent space, have been shown to be misspecified when fitting different classes of images. In particular, the generator will necessarily sample some low-quality images in between the classes. Rather than modifying the architecture, a line of works aims at improving the sampling quality from pre-trained generators at the expense of increased computational cost. Building on this, we introduce an additional network to predict latent importance weights and two associated sampling methods to avoid the poorest samples. This idea has several advantages: 1) it provides a way to inject disconnectedness into any GAN architecture, 2) since the rejection happens in the latent space, it avoids going through both the generator and the discriminator, saving computation time, 3) this importance weights formulation provides a principled way to reduce the Wasserstein's distance to the target distribution. We demonstrate the effectiveness of our method on several datasets, both synthetic and high-dimensional.