Abstract

To what extent can Variational Autoencoders (VAEs) identify semantically meaningful latent variables? Can they at least capture the correct topology if ground-truth latent variables are known? To investigate these questions, we introduce the Diffusion VAE, which allows for arbitrary (closed) manifolds in latent space. A Diffusion VAE uses transition kernels of Brownian motion on the manifold. In particular, it uses properties of the Brownian motion to implement the reparametrization trick and fast approximations to the KL divergence. We show that the Diffusion Variational Autoencoder is indeed capable of capturing topological properties.
Original languageEnglish
Publication statusPublished - 2019
EventNeurIPS 2019 Workshop - Vancouver Convention Center, Vancouver, Canada
Duration: 13 Dec 201913 Dec 2019

Conference

ConferenceNeurIPS 2019 Workshop
CountryCanada
CityVancouver
Period13/12/1913/12/19

Fingerprint Dive into the research topics of 'Can VAEs capture topological properties?'. Together they form a unique fingerprint.

  • Cite this

    Perez Rey, L. A., Menkovski, V., & Portegies, J. W. (2019). Can VAEs capture topological properties?. Paper presented at NeurIPS 2019 Workshop, Vancouver, Canada.