Multi-Level Variational Autoencoder: Learning Disentangled Representations from Grouped Observations

AAAI 2018 |

We would like to learn a representation of the data that reflects the semantics behind a specific grouping of the data, where within a group the samples share a common factor of variation. For example, consider a set of face images grouped by identity. We wish to anchor the semantics of the grouping into a disentangled representation that we can exploit. However, existing deep probabilistic models often assume that the samples are independent and identically distributed, thereby disregard the grouping information. We present the Multi-Level Variational Autoencoder (ML-VAE), a new deep probabilistic model for learning a disentangled representation of grouped data. The ML-VAE separates the latent representation into semantically relevant parts by working both at the group level and the observation level, while retaining efficient test-time inference. We experimentally show that our model (i) learns a semantically meaningful disentanglement, (ii) enables control over the latent representation, and (iii) generalises to unseen groups.