We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Mathematics > Statistics Theory

Title: Optimal estimation of Gaussian DAG models

Abstract: We study the optimal sample complexity of learning a Gaussian directed acyclic graph (DAG) from observational data. Our main results establish the minimax optimal sample complexity for learning the structure of a linear Gaussian DAG model in two settings of interest: 1) Under equal variances without knowledge of the true ordering, and 2) For general linear models given knowledge of the ordering. In both cases the sample complexity is $n\asymp q\log(d/q)$, where $q$ is the maximum number of parents and $d$ is the number of nodes. We further make comparisons with the classical problem of learning (undirected) Gaussian graphical models, showing that under the equal variance assumption, these two problems share the same optimal sample complexity. In other words, at least for Gaussian models with equal error variances, learning a directed graphical model is statistically no more difficult than learning an undirected graphical model. Our results also extend to more general identification assumptions as well as subgaussian errors.
Comments: 21 pages, 2 figures, to appear in AISTATS 2022
Subjects: Statistics Theory (math.ST); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Machine Learning (stat.ML)
Cite as: arXiv:2201.10548 [math.ST]
  (or arXiv:2201.10548v2 [math.ST] for this version)

Submission history

From: Bryon Aragam [view email]
[v1] Tue, 25 Jan 2022 18:56:56 GMT (81kb,D)
[v2] Sun, 20 Mar 2022 21:44:49 GMT (84kb,D)

Link back to: arXiv, form interface, contact.