Learnable Explicit Density for Continuous Latent Space and Variational Inference
This work addresses the problem of inflexible latent distributions in VAEs for researchers in generative modeling, offering incremental improvements in density estimation and inference.
The paper tackles the limitations of variational autoencoders (VAEs) by proposing a learnable explicit density approach for continuous latent spaces, enabling flexible priors and improved posterior approximations without relying on factorial Gaussians.
In this paper, we study two aspects of the variational autoencoder (VAE): the prior distribution over the latent variables and its corresponding posterior. First, we decompose the learning of VAEs into layerwise density estimation, and argue that having a flexible prior is beneficial to both sample generation and inference. Second, we analyze the family of inverse autoregressive flows (inverse AF) and show that with further improvement, inverse AF could be used as universal approximation to any complicated posterior. Our analysis results in a unified approach to parameterizing a VAE, without the need to restrict ourselves to use factorial Gaussians in the latent real space.