Intervening to learn and compose causally disentangled representations

Alex Markham, Isaac Hirsch, Jeri A. Chang, Liam Solus, Bryon Aragam

Keywords: causal disentanglement, generative models, intervention, composition, abstraction, causality

Abstract: In designing generative models, it is commonly believed that in order to learn useful latent structure, we face a fundamental tension between expressivity and structure. In this paper we challenge this view by proposing a new approach to training arbitrarily expressive generative models that simultaneously learn causally disentangled concepts. This is accomplished by adding a simple decoder-only module to an existing decoder that can be arbitrarily complex. The module learns to process concept information by implicitly inverting linear representations from an encoder. Inspired by the notion of intervention in a causal model, our module selectively modifies its architecture during training, allowing it to learn a compact joint model over different contexts. We show how adding this module leads to causally disentangled representations that can be composed for out-of-distribution generation on both real and simulated data. To further validate our proposed approach, we prove a new identifiability result that extends existing work on identifying structured representations in nonlinear models.

Primary Area: generative models

Submission Number: 13139