Weakly supervised causal representation learning
This addresses the challenge of causal inference in AI for domains like robotics, offering a novel approach but with incremental improvements in methodology.
The paper tackles the problem of learning causal representations from unstructured data like pixels, proving identifiability under weak supervision with paired pre- and post-intervention samples, and demonstrates on image data that their model reliably identifies causal structure and disentangles variables.
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 involves 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.