HM-VAEs: a Deep Generative Model for Real-valued Data with Heterogeneous Marginals

Chao Ma, Sebastian Tschiatschek, Yingzhen Li, Richard Turner, Jose Miguel Hernandez-Lobato, Cheng Zhang

Abstract: In this paper, we focused on improving VAEs for real-valued data that has heterogeneous marginal distributions. We propose the heterogeneous-marginal VAE (HM-VAE), a method that explicitly decomposes intra-variable uncertainties and inter-variable uncertainties. We experimentally observe that the HM-VAEs can generate realistic data with nearly indistinguishable marginals when compared with real data.