Arbitrarily-conditioned Data Imputation

Micael Carvalho, Thibaut Durand, Jiawei He, Nazanin Mehrasa, Greg Mori

Keywords: data imputation, variational autoencoder, normalizing flow

TL;DR: We propose an arbitrarily-conditioned data imputation framework built upon variational autoencoders and normalizing flows

Abstract: In this paper, we propose an arbitrarily-conditioned data imputation framework built upon variational autoencoders and normalizing flows. The proposed model is capable of mapping any partial data to a multi-modal latent variational distribution. Sampling from such a distribution leads to stochastic imputation. Preliminary evaluation on MNIST dataset shows promising stochastic imputation conditioned on partial images as input.