Learning discriminative Bayesian networks from high-dimensional continuous neuroimaging data

Luping Zhou, Lei Wang, Lingqiao Liu, Philip Ogunbona, Dinggang Shen

Research output: Contribution to journalArticle

12 Citations (Scopus)

Abstract

Due to its causal semantics, Bayesian networks (BN) have been widely employed to discover the underlying data relationship in exploratory studies, such as brain research. Despite its success in modeling the probability distribution of variables, BN is naturally a generative model, which is not necessarily discriminative. This may cause the ignorance of subtle but critical network changes that are of investigation values across populations. In this paper, we propose to improve the discriminative power of BN models for continuous variables from two different perspectives. This brings two general discriminative learning frameworks for Gaussian Bayesian networks (GBN). In the first framework, we employ Fisher kernel to bridge the generative models of GBN and the discriminative classifiers of SVMs, and convert the GBN parameter learning to Fisher kernel learning via minimizing a generalization error bound of SVMs. In the second framework, we employ the max-margin criterion and build it directly upon GBN models to explicitly optimize the classification performance of the GBNs. The advantages and disadvantages of the two frameworks are discussed and experimentally compared. Both of them demonstrate strong power in learning discriminative parameters of GBNs for neuroimaging based brain network analysis, as well as maintaining reasonable representation capacity. The contributions of this paper also include a new Directed Acyclic Graph (DAG) constraint with theoretical guarantee to ensure the graph validity of GBN.

Original languageEnglish
Article number7364252
Pages (from-to)2269-2283
Number of pages15
JournalIEEE Transactions on Pattern Analysis and Machine Intelligence
Volume38
Issue number11
DOIs
Publication statusPublished - 2016 Nov 1

Fingerprint

Neuroimaging
Bayesian networks
Bayesian Networks
High-dimensional
Learning
Generative Models
Brain
Bayesian Model
Semantics
Network Model
kernel
Generalization Error
Parameter Learning
Directed Acyclic Graph
Network Analysis
Continuous Variables
Electric network analysis
Research
Margin
Probability distributions

Keywords

  • Bayesian network
  • Brain network
  • Discriminative learning
  • Fisher kernel learning
  • Max-margin

ASJC Scopus subject areas

  • Software
  • Computer Vision and Pattern Recognition
  • Computational Theory and Mathematics
  • Artificial Intelligence
  • Applied Mathematics

Cite this

Learning discriminative Bayesian networks from high-dimensional continuous neuroimaging data. / Zhou, Luping; Wang, Lei; Liu, Lingqiao; Ogunbona, Philip; Shen, Dinggang.

In: IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 38, No. 11, 7364252, 01.11.2016, p. 2269-2283.

Research output: Contribution to journalArticle

@article{2250c74764b64cb48f2ca14820574ffb,
title = "Learning discriminative Bayesian networks from high-dimensional continuous neuroimaging data",
abstract = "Due to its causal semantics, Bayesian networks (BN) have been widely employed to discover the underlying data relationship in exploratory studies, such as brain research. Despite its success in modeling the probability distribution of variables, BN is naturally a generative model, which is not necessarily discriminative. This may cause the ignorance of subtle but critical network changes that are of investigation values across populations. In this paper, we propose to improve the discriminative power of BN models for continuous variables from two different perspectives. This brings two general discriminative learning frameworks for Gaussian Bayesian networks (GBN). In the first framework, we employ Fisher kernel to bridge the generative models of GBN and the discriminative classifiers of SVMs, and convert the GBN parameter learning to Fisher kernel learning via minimizing a generalization error bound of SVMs. In the second framework, we employ the max-margin criterion and build it directly upon GBN models to explicitly optimize the classification performance of the GBNs. The advantages and disadvantages of the two frameworks are discussed and experimentally compared. Both of them demonstrate strong power in learning discriminative parameters of GBNs for neuroimaging based brain network analysis, as well as maintaining reasonable representation capacity. The contributions of this paper also include a new Directed Acyclic Graph (DAG) constraint with theoretical guarantee to ensure the graph validity of GBN.",
keywords = "Bayesian network, Brain network, Discriminative learning, Fisher kernel learning, Max-margin",
author = "Luping Zhou and Lei Wang and Lingqiao Liu and Philip Ogunbona and Dinggang Shen",
year = "2016",
month = "11",
day = "1",
doi = "10.1109/TPAMI.2015.2511754",
language = "English",
volume = "38",
pages = "2269--2283",
journal = "IEEE Transactions on Pattern Analysis and Machine Intelligence",
issn = "0162-8828",
publisher = "IEEE Computer Society",
number = "11",

}

TY - JOUR

T1 - Learning discriminative Bayesian networks from high-dimensional continuous neuroimaging data

AU - Zhou, Luping

AU - Wang, Lei

AU - Liu, Lingqiao

AU - Ogunbona, Philip

AU - Shen, Dinggang

PY - 2016/11/1

Y1 - 2016/11/1

N2 - Due to its causal semantics, Bayesian networks (BN) have been widely employed to discover the underlying data relationship in exploratory studies, such as brain research. Despite its success in modeling the probability distribution of variables, BN is naturally a generative model, which is not necessarily discriminative. This may cause the ignorance of subtle but critical network changes that are of investigation values across populations. In this paper, we propose to improve the discriminative power of BN models for continuous variables from two different perspectives. This brings two general discriminative learning frameworks for Gaussian Bayesian networks (GBN). In the first framework, we employ Fisher kernel to bridge the generative models of GBN and the discriminative classifiers of SVMs, and convert the GBN parameter learning to Fisher kernel learning via minimizing a generalization error bound of SVMs. In the second framework, we employ the max-margin criterion and build it directly upon GBN models to explicitly optimize the classification performance of the GBNs. The advantages and disadvantages of the two frameworks are discussed and experimentally compared. Both of them demonstrate strong power in learning discriminative parameters of GBNs for neuroimaging based brain network analysis, as well as maintaining reasonable representation capacity. The contributions of this paper also include a new Directed Acyclic Graph (DAG) constraint with theoretical guarantee to ensure the graph validity of GBN.

AB - Due to its causal semantics, Bayesian networks (BN) have been widely employed to discover the underlying data relationship in exploratory studies, such as brain research. Despite its success in modeling the probability distribution of variables, BN is naturally a generative model, which is not necessarily discriminative. This may cause the ignorance of subtle but critical network changes that are of investigation values across populations. In this paper, we propose to improve the discriminative power of BN models for continuous variables from two different perspectives. This brings two general discriminative learning frameworks for Gaussian Bayesian networks (GBN). In the first framework, we employ Fisher kernel to bridge the generative models of GBN and the discriminative classifiers of SVMs, and convert the GBN parameter learning to Fisher kernel learning via minimizing a generalization error bound of SVMs. In the second framework, we employ the max-margin criterion and build it directly upon GBN models to explicitly optimize the classification performance of the GBNs. The advantages and disadvantages of the two frameworks are discussed and experimentally compared. Both of them demonstrate strong power in learning discriminative parameters of GBNs for neuroimaging based brain network analysis, as well as maintaining reasonable representation capacity. The contributions of this paper also include a new Directed Acyclic Graph (DAG) constraint with theoretical guarantee to ensure the graph validity of GBN.

KW - Bayesian network

KW - Brain network

KW - Discriminative learning

KW - Fisher kernel learning

KW - Max-margin

UR - http://www.scopus.com/inward/record.url?scp=84991693784&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84991693784&partnerID=8YFLogxK

U2 - 10.1109/TPAMI.2015.2511754

DO - 10.1109/TPAMI.2015.2511754

M3 - Article

AN - SCOPUS:84991693784

VL - 38

SP - 2269

EP - 2283

JO - IEEE Transactions on Pattern Analysis and Machine Intelligence

JF - IEEE Transactions on Pattern Analysis and Machine Intelligence

SN - 0162-8828

IS - 11

M1 - 7364252

ER -