Learning nonparametric latent causal graphs with unknown interventions

TL;DR

This paper proves that latent causal graphs can be uniquely identified from unknown interventions without parametric assumptions, extending causal representation learning to more general, nonparametric settings.

Contribution

It introduces new graphical concepts and conditions for nonparametric identifiability of latent causal structures with unknown interventions, without assuming known number of hidden variables.

Findings

Latent causal graphs are identifiable under certain conditions.

Only one unknown intervention per hidden variable is needed.

New graphical tools and characterizations of DAG equivalence classes are developed.

Abstract

We establish conditions under which latent causal graphs are nonparametrically identifiable and can be reconstructed from unknown interventions in the latent space. Our primary focus is the identification of the latent structure in measurement models without parametric assumptions such as linearity or Gaussianity. Moreover, we do not assume the number of hidden variables is known, and we show that at most one unknown intervention per hidden variable is needed. This extends a recent line of work on learning causal representations from observations and interventions. The proofs are constructive and introduce two new graphical concepts -- imaginary subsets and isolated edges -- that may be useful in their own right. As a matter of independent interest, the proofs also involve a novel characterization of the limits of edge orientations within the equivalence class of DAGs induced by unknown interventions. These are the first results to characterize the conditions under which causal representations are identifiable without making any parametric assumptions in a general setting with unknown interventions and without faithfulness.