Lower Bounds on the Size of Markov Equivalence Classes

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Jahn, Erik, Eberhardt, Frederick, Schulman, Leonard J.
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909705031385088
author Jahn, Erik
Eberhardt, Frederick
Schulman, Leonard J.
author_facet Jahn, Erik
Eberhardt, Frederick
Schulman, Leonard J.
contents Causal discovery algorithms typically recover causal graphs only up to their Markov equivalence classes unless additional parametric assumptions are made. The sizes of these equivalence classes reflect the limits of what can be learned about the underlying causal graph from purely observational data. Under the assumptions of acyclicity, causal sufficiency, and a uniform model prior, Markov equivalence classes are known to be small on average. In this paper, we show that this is no longer the case when any of these assumptions is relaxed. Specifically, we prove exponentially large lower bounds for the expected size of Markov equivalence classes in three settings: sparse random directed acyclic graphs, uniformly random acyclic directed mixed graphs, and uniformly random directed cyclic graphs.
format Preprint
id arxiv_https___arxiv_org_abs_2506_20933
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Lower Bounds on the Size of Markov Equivalence Classes
Jahn, Erik
Eberhardt, Frederick
Schulman, Leonard J.
Machine Learning
Statistics Theory
Causal discovery algorithms typically recover causal graphs only up to their Markov equivalence classes unless additional parametric assumptions are made. The sizes of these equivalence classes reflect the limits of what can be learned about the underlying causal graph from purely observational data. Under the assumptions of acyclicity, causal sufficiency, and a uniform model prior, Markov equivalence classes are known to be small on average. In this paper, we show that this is no longer the case when any of these assumptions is relaxed. Specifically, we prove exponentially large lower bounds for the expected size of Markov equivalence classes in three settings: sparse random directed acyclic graphs, uniformly random acyclic directed mixed graphs, and uniformly random directed cyclic graphs.
title Lower Bounds on the Size of Markov Equivalence Classes
topic Machine Learning
Statistics Theory
url https://arxiv.org/abs/2506.20933