A Geometric Perspective on Variational Autoencoders
This work addresses the challenge of enhancing generative model performance and interpolation quality for researchers and practitioners in machine learning, though it is incremental as it builds on existing VAE frameworks.
The paper tackles the problem of improving Variational Autoencoders by introducing a geometric perspective that reveals a Riemannian structure in the latent space, leading to a new sampling method that makes vanilla VAEs competitive or better than advanced versions on benchmark datasets, with demonstrated robustness in low-data regimes.
This paper introduces a new interpretation of the Variational Autoencoder framework by taking a fully geometric point of view. We argue that vanilla VAE models unveil naturally a Riemannian structure in their latent space and that taking into consideration those geometrical aspects can lead to better interpolations and an improved generation procedure. This new proposed sampling method consists in sampling from the uniform distribution deriving intrinsically from the learned Riemannian latent space and we show that using this scheme can make a vanilla VAE competitive and even better than more advanced versions on several benchmark datasets. Since generative models are known to be sensitive to the number of training samples we also stress the method's robustness in the low data regime.