Canada Spends Logo

Explore our Budget 2026 analysis

Grant: Discovery Grants Program - Individual

Simon Fraser University

$38,000.00

Bayesian Functional Data Analysis

FY 2024-25

Summary
Functional data analysis (FDA) is attracting increasing interest due to its ability to analyze curves, images, and multidimensional functions. Existing methodologies exhibit limitations in terms of scalability, interpretability, and uncertainty quantification. Bayesian FDA offers a flexible and interpretable approach for modeling functional data and provides uncertainty quantification for decision-making. However, challenges exist in the development of robust and dynamic Bayesian models, dimension reduction, and efficient online Bayesian inference for complex models with large-scale functional data. My long-term research goal is to develop novel statistical models and scalable computational methods for Bayesian FDA. I will establish foundational frameworks for modeling various large-scale functional data and develop computationally efficient methods with user-friendly software. Over the next 5 years, my research team will address 4 short-term objectives: 1) Develop scalable Monte Carlo methods for Bayesian inference with complex models and large data. The proposed method offers a unified framework that seamlessly integrates various state-of-the-art methods, providing parallel efficiency, unbiased estimators with variance for marginal likelihood, and adaptability for online inference. 2) Develop time-varying robust Bayesian functional principal component analysis (FPCA) for noisy and streaming data. Standard FPCA neglects serial dependence, leading to inadequate dimension reduction. I propose to leverage temporal dependence for improved dimension reduction and feature extraction. 3) Conduct robust Bayesian inference for large-scale dynamic models presented by nonlinear ordinary differential equations (ODEs). I propose to approximate the ODE solutions using a linear combination of basis functions to avoid expensive ODE solvers and achieve robustness. I will address the challenges in choosing prior distributions and updating inference with new data. 4) Create Bayesian models tailored for multidimensional functional data. This research will establish new models for a series of longitudinally observed multidimensional functional data, such as environmental data collected at various locations. I will develop a dynamic Bayesian FPCA method by estimating the time-varying covariance function, from which we will extract features to model and predict the specific predictor under investigation. The proposed research is innovative and will lead to groundbreaking advances in Bayesian inference and FDA. It can mitigate fundamental limitations of conventional methods for Bayesian inference while harnessing their collective strengths. We will develop publicly available software for Bayesian FDA, likely outperforming existing tools, such as WinBUGS and Stan. Our methods and software will have a broad impact on research and industry. The proposed research provides a wide variety of cutting-edge topics for training the next generation of statisticians.
Program
Discovery Grants Program - Individual
Location
British Columbia, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data

As published

Application id
RGPIN-2024-06839
Name nom
Wang, Liangliang
Department département
Statistical and Actuarial Sciences
Organization id
5
Institution établissement
Simon Fraser University
Province en
British Columbia
Province fr
Colombie-Britannique
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2024
Award amount
38000
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
1201
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Life sciences (including biotechnology)
Area of application fr
Sciences de la vie (y compris biotechnologie)
Research subject code
3000
Research subject group en
Statistics and probability
Research subject group fr
Statistique et probabilité
Research subject en
Statistics and probability
Research subject fr
Statistique et probabilité
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
Bayesian Functional Data Analysis
Keywords
functional data analysis; functional principal component analysis; bayesian functional data analysis; ordinary differential equations; markov chain monte carlo; sequential monte carlo
Application summary
Functional data analysis (FDA) is attracting increasing interest due to its ability to analyze curves, images, and multidimensional functions. Existing methodologies exhibit limitations in terms of scalability, interpretability, and uncertainty quantification. Bayesian FDA offers a flexible and interpretable approach for modeling functional data and provides uncertainty quantification for decision-making. However, challenges exist in the development of robust and dynamic Bayesian models, dimension reduction, and efficient online Bayesian inference for complex models with large-scale functional data. My long-term research goal is to develop novel statistical models and scalable computational methods for Bayesian FDA. I will establish foundational frameworks for modeling various large-scale functional data and develop computationally efficient methods with user-friendly software. Over the next 5 years, my research team will address 4 short-term objectives: 1) Develop scalable Monte Carlo methods for Bayesian inference with complex models and large data. The proposed method offers a unified framework that seamlessly integrates various state-of-the-art methods, providing parallel efficiency, unbiased estimators with variance for marginal likelihood, and adaptability for online inference. 2) Develop time-varying robust Bayesian functional principal component analysis (FPCA) for noisy and streaming data. Standard FPCA neglects serial dependence, leading to inadequate dimension reduction. I propose to leverage temporal dependence for improved dimension reduction and feature extraction. 3) Conduct robust Bayesian inference for large-scale dynamic models presented by nonlinear ordinary differential equations (ODEs). I propose to approximate the ODE solutions using a linear combination of basis functions to avoid expensive ODE solvers and achieve robustness. I will address the challenges in choosing prior distributions and updating inference with new data. 4) Create Bayesian models tailored for multidimensional functional data. This research will establish new models for a series of longitudinally observed multidimensional functional data, such as environmental data collected at various locations. I will develop a dynamic Bayesian FPCA method by estimating the time-varying covariance function, from which we will extract features to model and predict the specific predictor under investigation. The proposed research is innovative and will lead to groundbreaking advances in Bayesian inference and FDA. It can mitigate fundamental limitations of conventional methods for Bayesian inference while harnessing their collective strengths. We will develop publicly available software for Bayesian FDA, likely outperforming existing tools, such as WinBUGS and Stan. Our methods and software will have a broad impact on research and industry. The proposed research provides a wide variety of cutting-edge topics for training the next generation of statisticians.