Distributions Over DAGs for Causal Discovery: Limitations of Expressiveness

MCML Authors

Simon Rittel

→ Group Bernd Bischl
 Statistical Learning and Data Science

Abstract

Bayesian approaches for causal discovery can —inprinciple— quantify uncertainty in the prediction of the underlying causal structure, typically modeled by a directed acyclic graph (DAG). Various semi-implicit models for parametrized distributions over DAGs have been proposed, but their limitations have not been studied thoroughly. In this work, we focus on the expressiveness of parametrized distributions over DAGs in the context of causal discovery. We show several limitations of candidate models in a theoretical analysis and validate them empirically in supervised settings. To overcome these limitations, we propose using mixture models of the considered distributions over DAGs.

inproceedings RT25

---

MLG @ECML-PKDD 2025

22nd International Workshop on Mining and Learning with Graphs at European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases. Porto, Portugal, Sep 15-19, 2025.

---

Authors

S. Rittel • S. Tschiatschek

---

Links

PDF

---

Research Area

 A1 | Statistical Foundations & Explainability