Unobserved discrete data are ubiquitous in many scientific disciplines, and how to learn the causal structure of these latent variables is crucial for uncovering data patterns. Most studies focus on the linear latent variable model or impose strict constraints on latent structures, which fail to address cases in discrete data involving non-linear relationships or complex latent structures.

Existing methods can be broadly divided into two categories:importance sampling(IS) based and model-based [58].

A promising direction istouse linear recurrent architectures (LRUs), where dense recurrent weights are replaced with a complex-valued diagonal, making RTRL efficient.

However,macro-action design isoften heuristic and rarely comes with performance guarantees.