Grant: Discovery Grants Program - Individual
York University
$34,000.00
Statistical learning methods for multi-task and network data
FY 2024-25
Summary
Statistical Learning Methods for Network and Multi-Task Data
With the advent of technology, there is an increasing amount of network data that has complex dependency relationships. To model the dynamic feature of the network structures, we investigate new modelling strategies to model networks that changes with respect to different influencing factors. Another specific aim of the proposal is to develop statistical methods that integrates information from different data sources and constructs learning models for multiple tasks simultaneously. In this proposal, we aim to embrace the challenges and develop new statistical methods for network and multi-task data.
Project A: For high dimensional multivariate data, the underlying network can be influenced by some important factors. There is a lack of methods that model the relationship between the underlying network structure and the influential covariates. We propose to develop a covariate dependent Gaussian graphic model, in which the underlying graphic model parameters are functionally dependent on the covariates. Likelihood or composite likelihood formulation can be used to characterize the underlying relationship between the graphic model and the covariates. For high dimensional graphical models, we can impose sparsity on the model by introducing regularization on the model. We aim to establish estimation consistency and model selection consistency for the proposed estimators. To further reduce the dimensionality of the model, we can introduce symmetry constraints to the underlying covariate varied Gaussian graphical model. We propose colored covariate varied Gaussian graphical model to include the equality constraints on the model parameters. Such model has a underlying color structure on the vertices and edges and also allows the magnitude of edges to change with respect to the covariates.
Project B: We will develop new Bayesian based methods for the analysis of covariate dependent Gaussian graphic model. Our goals include the development of estimation methods for variable dependent GGM via MCMC sampling strategies. We also aim to develop efficient MCMC algorithms to sample high dimensional covariate dependent GGM and perform global model selections over the model space and extend the methods to color constrained covariate dependent Gaussian graphic model.
Project C:Data integration involves the simultaneous analysis of multiple interconnected data sets collected from diverse sources. We propose to develop data integration methods which utilize working correlation to combine correlated tasks and perform feature learning. We will extend our approach to perform high dimensional inference using information cross multiple platforms. We will consider high dimensional inference methods and use de-correlation strategy to construct test statistics which are able to test linear hypothesis on high dimensional parameters in multi-task setting.
I
Program
Discovery Grants Program - Individual
Recipient
Location
Ontario, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data
As published
Application id
RGPIN-2024-06202
Name nom
Gao, Xin
Department département
Mathematics and Statistics
Organization id
38
Institution établissement
York University
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2024
Award amount
34000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
DISCOVERY RESEARCH
Group fr
RECHERCHE AXÉE SUR LA DÉCOUVERTE
Committee code
1508
Committee name en
Mathematics and Statistics
Committee name fr
Mathématiques et statistique
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
3001
Research subject group en
Statistics and probability
Research subject group fr
Statistique et probabilité
Research subject en
Statistical theory
Research subject fr
Théorie de statistique
Field of research list names en
Not available
Field of research list names fr
Non disponible
Socioeconomic objective list names en
Not available
Socioeconomic objective list names fr
Non disponible
Application title
Statistical learning methods for multi-task and network data
Keywords
network modeling; multi-task learning; data integration; dynamic graphic model; colored graphic model; heterogeneous data source; artificial intelligence for medical studies; covariate dependent network; integration of language model in biostatistics; integration of multiomics data
Application summary
Statistical Learning Methods for Network and Multi-Task Data
With the advent of technology, there is an increasing amount of network data that has complex dependency relationships. To model the dynamic feature of the network structures, we investigate new modelling strategies to model networks that changes with respect to different influencing factors. Another specific aim of the proposal is to develop statistical methods that integrates information from different data sources and constructs learning models for multiple tasks simultaneously. In this proposal, we aim to embrace the challenges and develop new statistical methods for network and multi-task data.
Project A: For high dimensional multivariate data, the underlying network can be influenced by some important factors. There is a lack of methods that model the relationship between the underlying network structure and the influential covariates. We propose to develop a covariate dependent Gaussian graphic model, in which the underlying graphic model parameters are functionally dependent on the covariates. Likelihood or composite likelihood formulation can be used to characterize the underlying relationship between the graphic model and the covariates. For high dimensional graphical models, we can impose sparsity on the model by introducing regularization on the model. We aim to establish estimation consistency and model selection consistency for the proposed estimators. To further reduce the dimensionality of the model, we can introduce symmetry constraints to the underlying covariate varied Gaussian graphical model. We propose colored covariate varied Gaussian graphical model to include the equality constraints on the model parameters. Such model has a underlying color structure on the vertices and edges and also allows the magnitude of edges to change with respect to the covariates.
Project B: We will develop new Bayesian based methods for the analysis of covariate dependent Gaussian graphic model. Our goals include the development of estimation methods for variable dependent GGM via MCMC sampling strategies. We also aim to develop efficient MCMC algorithms to sample high dimensional covariate dependent GGM and perform global model selections over the model space and extend the methods to color constrained covariate dependent Gaussian graphic model.
Project C:Data integration involves the simultaneous analysis of multiple interconnected data sets collected from diverse sources. We propose to develop data integration methods which utilize working correlation to combine correlated tasks and perform feature learning. We will extend our approach to perform high dimensional inference using information cross multiple platforms. We will consider high dimensional inference methods and use de-correlation strategy to construct test statistics which are able to test linear hypothesis on high dimensional parameters in multi-task setting.
I