Weakly supervised causal representation learning

Johann Brehmer, Pim De Haan, Phillip Lippe, Taco Cohen

Keywords: causal representation learning, causality, disentangled representation learning

TL;DR: We show that causal factors and their causal structure can be identified from low-level data (e.g. pixels) observed before and after interventions.

Abstract: Learning high-level causal representations together with a causal model from unstructured low-level data such as pixels is impossible from observational data alone. We prove under mild assumptions that this representation is however identifiable in a weakly supervised setting. This requires a dataset with paired samples before and after random, unknown interventions, but no further labels. We then introduce implicit latent causal models, variational autoencoders that represent causal variables and causal structure without having to optimize an explicit discrete graph structure. On simple image data, including a novel dataset of simulated robotic manipulation, we demonstrate that such models can reliably identify the causal structure and disentangle causal variables.