Get our free extension to see links to code for papers anywhere online!

Chrome logo Add to Chrome

Firefox logo Add to Firefox


Hyperspherical Variational Auto-Encoders

Sep 26, 2018
Tim R. Davidson, Luca Falorsi, Nicola De Cao, Thomas Kipf, Jakub M. Tomczak



The Variational Auto-Encoder (VAE) is one of the most used unsupervised machine learning models. But although the default choice of a Gaussian distribution for both the prior and posterior represents a mathematically convenient distribution often leading to competitive results, we show that this parameterization fails to model data with a latent hyperspherical structure. To address this issue we propose using a von Mises-Fisher (vMF) distribution instead, leading to a hyperspherical latent space. Through a series of experiments we show how such a hyperspherical VAE, or $\mathcal{S}$-VAE, is more suitable for capturing data with a hyperspherical latent structure, while outperforming a normal, $\mathcal{N}$-VAE, in low dimensions on other data types.

* Uncertainty in Artificial Intelligence (UAI). Proceedings of the Thirty-Fourth Conference (2018) 856- 865 
* GitHub repository: http://github.com/nicola-decao/s-vae-tf, Blogpost: https://nicola-decao.github.io/s-vae 


Share this with someone who'll enjoy it:

   Access Paper Source



Share this with someone who'll enjoy it: