Grant: Discovery Launch Supplement
University of Calgary
$12,500.00
New Machine Learning Methods in Functional Data Analysis
FY 2024-25
Summary
Functional data analysis is a branch of statistics that deals with curves, images, high-dimensional functions, or any complex form of these. Each random function is a sample element of function data that are common in applications using brain images, longitudinal data, and genetic data.
The long-term research vision is to establish machine learning-based and modern methodologies to address the increasing scale (in size or dimension) and complexity of functional data.
My first short-term objective is to establish a machine learning-based methodology for tackling the function-on-function regression problem and its variations. This method will incorporate a representation learning step that compresses and extracts information from original functional data into lower-dimensional representations. Using representations is more convenient for building prediction models than using original functional data.
My second short-term objective is to propose a nonlinear prediction model for panel functional time series. Panel functional time series is more complex where we could observe a panel of series simultaneously at different arbitrary locations. We will propose a nonlinear forecast model which is completely built with neural networks. It will be able to forecast the panel all at once and have potential applications in public health studies and financial market predictions.
My third short-term objective is to propose a generative model for functional data. Current generative models usually focus on discrete data, e.g., images defined on grids of pixels or sequences of words. These discrete data are produced based on some underlying continuous functions. It is then more beneficial to pursue a generative model for these underlying continuous signals.
My fourth short-term objective is to build an estimation tool for covariance functions and their associated eigenfunctions of multivariate functional data. I will utilize neural networks, as universal function approximators, to obtain covariance functions of multivariate functional data. This will improve the scalability of functional principal component analysis (FPCA) and make FPCA more applicable to multivariate functional data, e.g., brain images, longitudinal images, and spatial-temporal data.
The expected significance of the work is the innovation of machine learning methods in functional data analysis. The proposed research is likely to establish a new chapter for functional data analysis. It will lead to significant and ground-breaking advances in statistics. Furthermore, the proposed research vision produces new collaborative opportunities with other scientific disciplines and industry in Canada. We will publish user-friendly software for our research works to make them accessible to a broad audience. The proposed research provides a wide variety of cutting-edge topics and projects for highly qualified students and researchers.
Program
Discovery Launch Supplement
Recipient
Location
Alberta, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data
As published
Application id
DGECR-2024-00530
Name nom
Wang, Haixu
Department département
Mathematics and Statistics
Organization id
11
Institution établissement
University of Calgary
Province en
Alberta
Province fr
Alberta
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2024
Award amount
12500
Program id
DGECR
Program name en
Discovery Launch Supplement
Program name fr
Supplément Tremplin vers la découverte
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
1002
Area of application group en
Northern development
Area of application group fr
Développement du nord
Area of application en
Environment
Area of application fr
Environnement
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
New Machine Learning Methods in Functional Data Analysis
Keywords
functional data analysis; machine learning; functional time series analysis; generative models; longitudinal data analysis
Application summary
Functional data analysis is a branch of statistics that deals with curves, images, high-dimensional functions, or any complex form of these. Each random function is a sample element of function data that are common in applications using brain images, longitudinal data, and genetic data.
The long-term research vision is to establish machine learning-based and modern methodologies to address the increasing scale (in size or dimension) and complexity of functional data.
My first short-term objective is to establish a machine learning-based methodology for tackling the function-on-function regression problem and its variations. This method will incorporate a representation learning step that compresses and extracts information from original functional data into lower-dimensional representations. Using representations is more convenient for building prediction models than using original functional data.
My second short-term objective is to propose a nonlinear prediction model for panel functional time series. Panel functional time series is more complex where we could observe a panel of series simultaneously at different arbitrary locations. We will propose a nonlinear forecast model which is completely built with neural networks. It will be able to forecast the panel all at once and have potential applications in public health studies and financial market predictions.
My third short-term objective is to propose a generative model for functional data. Current generative models usually focus on discrete data, e.g., images defined on grids of pixels or sequences of words. These discrete data are produced based on some underlying continuous functions. It is then more beneficial to pursue a generative model for these underlying continuous signals.
My fourth short-term objective is to build an estimation tool for covariance functions and their associated eigenfunctions of multivariate functional data. I will utilize neural networks, as universal function approximators, to obtain covariance functions of multivariate functional data. This will improve the scalability of functional principal component analysis (FPCA) and make FPCA more applicable to multivariate functional data, e.g., brain images, longitudinal images, and spatial-temporal data.
The expected significance of the work is the innovation of machine learning methods in functional data analysis. The proposed research is likely to establish a new chapter for functional data analysis. It will lead to significant and ground-breaking advances in statistics. Furthermore, the proposed research vision produces new collaborative opportunities with other scientific disciplines and industry in Canada. We will publish user-friendly software for our research works to make them accessible to a broad audience. The proposed research provides a wide variety of cutting-edge topics and projects for highly qualified students and researchers.