← Search

Dimitrios Kalatzis

2 accepted papers

2022

Pulling back information geometry

AISTATS 2022poster

Latent space geometry has shown itself to provide a rich and rigorous framework for interacting with the latent variables of deep generative models. The existing theory, however, relies on the decoder being a Gaussian distribution as its simple reparametrization allows us to interpret the generating…

2020

Variational Autoencoders with Riemannian Brownian Motion Priors

ICML 2020poster

Variational Autoencoders (VAEs) represent the given data in a low-dimensional latent space, which is generally assumed to be Euclidean. This assumption naturally leads to the common choice of a standard Gaussian prior over continuous latent variables. Recent work has, however, shown that this prior…

Cited by 53SourcePDFScholar