Causal structure learning in directed, possibly cyclic, graphical models

We consider the problem of learning a directed graph G⋆{G}^{\star } from observational data. We assume that the distribution that gives rise to the samples is Markov and faithful to the graph G⋆{G}^{\star } and that there are no unobserved variables. We do not rely on any further assumptions regardi...

Full description

Saved in:
Bibliographic Details
Main Authors: Semnani Pardis, Robeva Elina
Format: Article
Language:English
Published: De Gruyter 2025-04-01
Series:Journal of Causal Inference
Subjects:
Online Access:https://doi.org/10.1515/jci-2024-0037
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850265764523671552
author Semnani Pardis
Robeva Elina
author_facet Semnani Pardis
Robeva Elina
author_sort Semnani Pardis
collection DOAJ
description We consider the problem of learning a directed graph G⋆{G}^{\star } from observational data. We assume that the distribution that gives rise to the samples is Markov and faithful to the graph G⋆{G}^{\star } and that there are no unobserved variables. We do not rely on any further assumptions regarding the graph or the distribution of the variables. Particularly, we allow for directed cycles in G⋆{G}^{\star } and work in the fully nonparametric setting. Given the set of conditional independence statements satisfied by the distribution, we aim to find a directed graph, which satisfies the same dd-separation statements as G⋆{G}^{\star }. We propose a hybrid approach consisting of two steps. We first find a partially ordered partition of the vertices of G⋆{G}^{\star } by optimizing a certain score in a greedy fashion. We prove that any optimal partition uniquely characterizes the Markov equivalence class of G⋆{G}^{\star }. Given an optimal partition, we propose an algorithm for constructing a graph in the Markov equivalence class of G⋆{G}^{\star } whose strongly connected components correspond to the elements of the partition, and which are partially ordered according to the partial order of the partition. Our algorithm comes in two versions – one that is provably correct and another one that performs fast in practice.
format Article
id doaj-art-59525e3512464a389bd4209f4cbdd044
institution OA Journals
issn 2193-3685
language English
publishDate 2025-04-01
publisher De Gruyter
record_format Article
series Journal of Causal Inference
spelling doaj-art-59525e3512464a389bd4209f4cbdd0442025-08-20T01:54:19ZengDe GruyterJournal of Causal Inference2193-36852025-04-011316012010.1515/jci-2024-0037Causal structure learning in directed, possibly cyclic, graphical modelsSemnani Pardis0Robeva Elina1Department of Mathematics, University of British Columbia, Vancouver, BC V6T 1Z2, CanadaDepartment of Mathematics, University of British Columbia, Vancouver, BC V6T 1Z2, CanadaWe consider the problem of learning a directed graph G⋆{G}^{\star } from observational data. We assume that the distribution that gives rise to the samples is Markov and faithful to the graph G⋆{G}^{\star } and that there are no unobserved variables. We do not rely on any further assumptions regarding the graph or the distribution of the variables. Particularly, we allow for directed cycles in G⋆{G}^{\star } and work in the fully nonparametric setting. Given the set of conditional independence statements satisfied by the distribution, we aim to find a directed graph, which satisfies the same dd-separation statements as G⋆{G}^{\star }. We propose a hybrid approach consisting of two steps. We first find a partially ordered partition of the vertices of G⋆{G}^{\star } by optimizing a certain score in a greedy fashion. We prove that any optimal partition uniquely characterizes the Markov equivalence class of G⋆{G}^{\star }. Given an optimal partition, we propose an algorithm for constructing a graph in the Markov equivalence class of G⋆{G}^{\star } whose strongly connected components correspond to the elements of the partition, and which are partially ordered according to the partial order of the partition. Our algorithm comes in two versions – one that is provably correct and another one that performs fast in practice.https://doi.org/10.1515/jci-2024-0037causal discoverydirected cyclic graphsmarkov equivalencefaithfulnessd-separationsconditional independence62d2062h2205c20
spellingShingle Semnani Pardis
Robeva Elina
Causal structure learning in directed, possibly cyclic, graphical models
Journal of Causal Inference
causal discovery
directed cyclic graphs
markov equivalence
faithfulness
d-separations
conditional independence
62d20
62h22
05c20
title Causal structure learning in directed, possibly cyclic, graphical models
title_full Causal structure learning in directed, possibly cyclic, graphical models
title_fullStr Causal structure learning in directed, possibly cyclic, graphical models
title_full_unstemmed Causal structure learning in directed, possibly cyclic, graphical models
title_short Causal structure learning in directed, possibly cyclic, graphical models
title_sort causal structure learning in directed possibly cyclic graphical models
topic causal discovery
directed cyclic graphs
markov equivalence
faithfulness
d-separations
conditional independence
62d20
62h22
05c20
url https://doi.org/10.1515/jci-2024-0037
work_keys_str_mv AT semnanipardis causalstructurelearningindirectedpossiblycyclicgraphicalmodels
AT robevaelina causalstructurelearningindirectedpossiblycyclicgraphicalmodels