Canada Spends Logo

Explorez notre analyse du budget 2026

Grant: Alliance Grants

The University of British Columbia

$25,000.00

Gaussian graphical model estimation - theory and applications

EF 2023-24

Résumé
Nowadays, the large amount of high-throughput data in fields such as genomics, finance and marketing poses a challenge to classical multivariate analysis. In particular, the number of samples is significantly smaller than the number of parameters. Gaussian graphical models (GGM) have proven to be an important tool for analyzing complex conditional independence relationships in high-dimensional data.To estimate the parameters in GGM, the Bayesian approach is particularly suitable for large-scale networks because it promotes graph sparsity, incorporates prior information, and, most importantly, takes into account uncertainty in the graph structure [1]. However, for two graphs differing by one edge, the ratio of approximate normalization constants has been a bottleneck in finding efficient model selection. Markov chain Monte Carlo (MCMC) methods are usually superior compared to other stochastic search techniques [2]. However, in high-dimensional spaces, MCMC requires significant computational time and has limited efficiency. A new method was developed in [3], which performs well in sparse graphs but may fail in the non-sparse graph case.To address the problem of estimation accuracy of the normalization constant, researchers from Canada and Qiong Li will apply the Gaussian Variation Approximation (GVA) method [4] based on their recent research. The expertise of our collaborators will facilitate our theoretical development, exploration of a wide range of applications, and provide an immersive cross-cultural learning experience for HQP training in Canada, in particular by imposing symmetric constraints on the inverse matrix of the covariance matrix to achieve colored edges [5,6,7]. The applicant's contribution will lead to further exploration of other efficient approximation methods such as the Laplace approximation and saddle-point approximation, paving the way for exciting new developments in this key area.
Programme
Alliance Grants
Emplacement
British Columbia, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data

Tel que publié

Application id
590341-2023
Name nom
Shi, Xiaoping X
Department département
Okanagan - Computer Science, Mathematics, Physics and Statistics
Organization id
2
Institution établissement
The University of British Columbia
Province en
British Columbia
Province fr
Colombie-Britannique
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2023
Competition year année de concours
2023
Award amount
25000
Program id
ALLRP
Program name en
Alliance Grants
Program name fr
Subventions Alliance
Group en
RESEARCH PARTNERSHIPS
Group fr
PARTENARIATS DE RECHERCHE
Committee code
96
Committee name en
RPP Internal Decision Cttee
Committee name fr
Cté de décision interne - PPR
Area of application code
1205
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Mathematical sciences
Area of application fr
Sciences mathématiques
Research subject code
3000
Research subject group en
Statistics and probability
Research subject group fr
Statistics and probability
Research subject en
Statistique et probabilité
Research subject fr
Statistique et probabilité
Application title
Gaussian graphical model estimation - theory and applications
Keyword
Colored Gaussian graphical models; Gaussian Variational approximation; High-dimensional analysis; Laplace approximation; Markov chain Monte Carlo; Multivariate analysis; Normalizing constants; Saddlepoint approximation
Application summary
Nowadays, the large amount of high-throughput data in fields such as genomics, finance and marketing poses a challenge to classical multivariate analysis. In particular, the number of samples is significantly smaller than the number of parameters. Gaussian graphical models (GGM) have proven to be an important tool for analyzing complex conditional independence relationships in high-dimensional data.To estimate the parameters in GGM, the Bayesian approach is particularly suitable for large-scale networks because it promotes graph sparsity, incorporates prior information, and, most importantly, takes into account uncertainty in the graph structure [1]. However, for two graphs differing by one edge, the ratio of approximate normalization constants has been a bottleneck in finding efficient model selection. Markov chain Monte Carlo (MCMC) methods are usually superior compared to other stochastic search techniques [2]. However, in high-dimensional spaces, MCMC requires significant computational time and has limited efficiency. A new method was developed in [3], which performs well in sparse graphs but may fail in the non-sparse graph case.To address the problem of estimation accuracy of the normalization constant, researchers from Canada and Qiong Li will apply the Gaussian Variation Approximation (GVA) method [4] based on their recent research. The expertise of our collaborators will facilitate our theoretical development, exploration of a wide range of applications, and provide an immersive cross-cultural learning experience for HQP training in Canada, in particular by imposing symmetric constraints on the inverse matrix of the covariance matrix to achieve colored edges [5,6,7]. The applicant's contribution will lead to further exploration of other efficient approximation methods such as the Laplace approximation and saddle-point approximation, paving the way for exciting new developments in this key area.