Object-centric architectures enable efficient causal representation learning
This addresses a fundamental limitation in causal representation learning for multi-object scenarios, though it appears incremental as it builds on existing Slot Attention and perturbation-based methods.
The paper tackles the problem of causal representation learning failing when observations contain multiple objects, by developing an object-centric architecture that combines Slot Attention with weak supervision from sparse perturbations. The result is a more data-efficient approach requiring significantly fewer perturbations than comparable Euclidean encoding methods, successfully disentangling object properties in simple image-based experiments.
Causal representation learning has showed a variety of settings in which we can disentangle latent variables with identifiability guarantees (up to some reasonable equivalence class). Common to all of these approaches is the assumption that (1) the latent variables are represented as $d$-dimensional vectors, and (2) that the observations are the output of some injective generative function of these latent variables. While these assumptions appear benign, we show that when the observations are of multiple objects, the generative function is no longer injective and disentanglement fails in practice. We can address this failure by combining recent developments in object-centric learning and causal representation learning. By modifying the Slot Attention architecture arXiv:2006.15055, we develop an object-centric architecture that leverages weak supervision from sparse perturbations to disentangle each object's properties. This approach is more data-efficient in the sense that it requires significantly fewer perturbations than a comparable approach that encodes to a Euclidean space and we show that this approach successfully disentangles the properties of a set of objects in a series of simple image-based disentanglement experiments.