Coarsening Linear Non-Gaussian Causal Models with Cycles
Francisco Madaleno, Francisco C Pereira, Alex Markham · 2026
Read Coarsening Linear Non-Gaussian Causal Models with Cycles by Francisco Madaleno, Francisco C Pereira, Alex Markham (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-10163.txt.
Description
Madaleno, Pereira and Markham extend causal abstraction to models with cycles. Recent graphical methods sum up a high-dimensional causal structure as a low-dimensional one by grouping variables into clusters. But they assume both levels are acyclic, which excludes many realistic models. In the linear non-Gaussian setting, the authors show the high-dimensional acyclicity assumption can be loosened while still getting back a low-dimensional causal directed acyclic graph.
Identifiability of that summary connects to known results. Cyclic linear non-Gaussian models are identifiable just up to an equivalence class whose members vary by reversals of directed cycles. The low-dimensional graph is invariant across the whole class, making it a natural stand-in. Learning the full equivalence class over high-dimensional variables takes exponential time. The low-dimensional summary is learned in cubic worst-case time with explicit sample limits, backed by open code and simulated tests.
Catalog record
| CREATOR | Francisco Madaleno, Francisco C Pereira, Alex Markham |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 54 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.10163