東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

FindBook

Google Book

Amazon

博客來

Structural Learning of Gaussian DAGs from Network Data.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Structural Learning of Gaussian DAGs from Network Data./
作者:	Li, Hangjian.
出版者:	Ann Arbor : ProQuest Dissertations & Theses, : 2021,
面頁冊數:	116 p.
附註:	Source: Dissertations Abstracts International, Volume: 83-03, Section: B.
Contained By:	Dissertations Abstracts International83-03B.
標題:	Statistics. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=28722152
ISBN:	9798544294634

Structural Learning of Gaussian DAGs from Network Data.
Li, Hangjian.

Structural Learning of Gaussian DAGs from Network Data. - Ann Arbor : ProQuest Dissertations & Theses, 2021 - 116 p.

Source: Dissertations Abstracts International, Volume: 83-03, Section: B.

Thesis (Ph.D.)--University of California, Los Angeles, 2021.

This item must not be sold to any third party vendors.

Structural learning of Gaussian directed acyclic graphs (DAGs) or Bayesian networks has been studied extensively under the assumption that data are independent. But in real applications such as in biology and social studies observations generated from a Bayesian network model are often mutually dependent and their dependence can be model by a second network model. In this dissertation, we generalize the existing Gaussian DAG framework by proposing a new Gaussian DAG model for dependent data which assumes the observations are correlated according to a given undirected network. Under this model, the dependent observations jointly follow a matrix normal distribution with variance represented by the Kronecker product of two positive definite matrices. The Cholesky factor of one of the matrices represent the DAG structure in the feature space while the other encodes the conditional independencies among the observations. We show that the proposed model also satisfies the desired score-equivalence property under common likelihood-based score functions. Based on the proposed model, we develop a block coordinate descent algorithm to estimate the DAG structure given a topological ordering of the vertices. The proposed algorithm jointly estimates a sparse Bayesian network and the correlations among observations by optimizing a scoring function based on penalized likelihood. The algorithm is fast and can scale to networks with thousands of nodes. We also established finite-sample error bounds and large-sample consistency of the estimators. In particular, we show that under some mild conditions, the proposed method produces consistent estimators for the DAG structure and the sample covariances after one iteration. Extensive numerical experiments also demonstrate that by jointly estimating the DAG structure and the sample correlation, our method achieves much higher accuracy in structure learning than the competing algorithms. When the node ordering is unknown, through experiments on synthetic and real data, we show that our algorithm can be used to estimate the correlations between samples, with which we can de-correlate the dependent data to significantly improve the performance of classical DAG learning methods.

ISBN: 9798544294634Subjects--Topical Terms:

517247
Statistics.
Subjects--Index Terms:

Bayesian networks

Structural Learning of Gaussian DAGs from Network Data.
LDR:03363nmm a2200373 4500 001 2346901
005 20220706051319.5
008 241004s2021 ||||||||||||||||| ||eng d
020 $a 9798544294634
035 $a (MiAaPQ)AAI28722152
035 $a AAI28722152
040 $a MiAaPQ $c MiAaPQ
100 1 $a Li, Hangjian. $3 3686100
245 1 0 $a Structural Learning of Gaussian DAGs from Network Data.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2021
300 $a 116 p.
500 $a Source: Dissertations Abstracts International, Volume: 83-03, Section: B.
500 $a Advisor: Zhou, Qing.
502 $a Thesis (Ph.D.)--University of California, Los Angeles, 2021.
506 $a This item must not be sold to any third party vendors.
520 $a Structural learning of Gaussian directed acyclic graphs (DAGs) or Bayesian networks has been studied extensively under the assumption that data are independent. But in real applications such as in biology and social studies observations generated from a Bayesian network model are often mutually dependent and their dependence can be model by a second network model. In this dissertation, we generalize the existing Gaussian DAG framework by proposing a new Gaussian DAG model for dependent data which assumes the observations are correlated according to a given undirected network. Under this model, the dependent observations jointly follow a matrix normal distribution with variance represented by the Kronecker product of two positive definite matrices. The Cholesky factor of one of the matrices represent the DAG structure in the feature space while the other encodes the conditional independencies among the observations. We show that the proposed model also satisfies the desired score-equivalence property under common likelihood-based score functions. Based on the proposed model, we develop a block coordinate descent algorithm to estimate the DAG structure given a topological ordering of the vertices. The proposed algorithm jointly estimates a sparse Bayesian network and the correlations among observations by optimizing a scoring function based on penalized likelihood. The algorithm is fast and can scale to networks with thousands of nodes. We also established finite-sample error bounds and large-sample consistency of the estimators. In particular, we show that under some mild conditions, the proposed method produces consistent estimators for the DAG structure and the sample covariances after one iteration. Extensive numerical experiments also demonstrate that by jointly estimating the DAG structure and the sample correlation, our method achieves much higher accuracy in structure learning than the competing algorithms. When the node ordering is unknown, through experiments on synthetic and real data, we show that our algorithm can be used to estimate the correlations between samples, with which we can de-correlate the dependent data to significantly improve the performance of classical DAG learning methods.
590 $a School code: 0031.
650 4 $a Statistics. $3 517247
650 4 $a Statistical physics. $3 536281
650 4 $a Information technology. $3 532993
650 4 $a Information science. $3 554358
653 $a Bayesian networks
653 $a Lasso
653 $a Matrix normal
653 $a Network data
690 $a 0463
690 $a 0489
690 $a 0723
690 $a 0217
710 2 $a University of California, Los Angeles. $b Statistics 0891. $3 2095317
773 0 $t Dissertations Abstracts International $g 83-03B.
790 $a 0031
791 $a Ph.D.
792 $a 2021
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=28722152