Sampling Generative Networks

TL;DR

This paper presents new techniques for sampling, visualizing, and analyzing the latent spaces of generative models like VAEs and GANs, improving sample quality and interpretability.

Contribution

It introduces spherical linear interpolation, J-Diagrams, MINE grids, and new attribute vector derivation methods, enhancing latent space exploration and analysis.

Findings

Spherical interpolation produces sharper samples.

J-Diagrams and MINE grids effectively visualize latent manifolds.

Attribute vectors enable quantitative latent space analysis.

Abstract

We introduce several techniques for sampling and visualizing the latent spaces of generative models. Replacing linear interpolation with spherical linear interpolation prevents diverging from a model's prior distribution and produces sharper samples. J-Diagrams and MINE grids are introduced as visualizations of manifolds created by analogies and nearest neighbors. We demonstrate two new techniques for deriving attribute vectors: bias-corrected vectors with data replication and synthetic vectors with data augmentation. Binary classification using attribute vectors is presented as a technique supporting quantitative analysis of the latent space. Most techniques are intended to be independent of model type and examples are shown on both Variational Autoencoders and Generative Adversarial Networks.