Canada Spends Logo

Explore our Budget 2026 analysis

Grant: Discovery Grants Program - Individual

University of Toronto

$30,000.00

Classification under High-dimensional Latent Factor Models

FY 2023-24

Summary
This proposal concerns the problem of classification, an important topic of statistical learning theory. High-dimensional latent factor models that have a low-dimensional, hidden structure, will be formulated to guarantee successful statistical classification performance based on suitable projections of the high-dimensional data. This project offers a principled way to analyze such high-dimensional datasets and will provide computationally efficient classification rules. The properties of our classifiers will be contrasted with others, including deep neural nets and interpolating classifiers. Results of this research are also expected to advance understanding on how to achieve optimal classification. Our proposed systematic treatment is a timely addition to the methodological and limited theoretical results in this important area. This project has important applications to recent advances in immunology and cancer studies, which revealed that hidden mechanisms can be directly connected to health outcomes. Specifically, this project constructs novel classifiers based on principal component analysis with a necessary debiasing part followed by linear discriminant analysis and develops their statistical and computational properties. In addition, this research aims to develop minimax adaptive bounds for the excess misclassification error of any classifier under general latent factor model specifications and to prove that the new methods achieve these bounds, thereby establishing their rate optimality. This project also studies of the important overparametrized classifiers that interpolate the data, but still possess good classification power, as well as the classifiers that base on deep neural nets after projection. Finally, the usefulness of the new techniques will be demonstrated via applications to data from immunology, neuroscience, economics, genomics and proteomics. This endeavor requires the close cooperation of an interdisciplinary team of researchers and graduate students at the University of Toronto. The proposed research will nurture the close ties between the PI and his scientific collaborators in Statistics and Data Sciences at Cornell University and Computational Biology and Immunology at the University of Pittsburgh. The results will be reported to the statistical, machine learning and computational biology communities at conferences, in reports made available on the world wide web, and in papers that will be submitted to leading statistical, machine learning and computational biology journals. Free software that implements the developed methodology will be made available on the web in a readily implementable form. Special effort will be made to involve undergraduates in parts of this research, as well as attracting master and Ph.D. students. Outreach to high school students and teachers will be done through the outreach programs at the Univeristy of Toronto with solid partnership with local schools, promoting equity and inclusion.
Program
Discovery Grants Program - Individual
Location
Ontario, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data

As published

Application id
RGPIN-2023-05207
Name nom
Bing, Xin
Department département
Department of Statistical Sciences
Organization id
31
Institution établissement
University of Toronto
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2023
Competition year année de concours
2023
Award amount
30000
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
9999
Research subject group en
Not available
Research subject group fr
Not available
Research subject en
Non disponible
Research subject fr
Non disponible
Application title
Classification under High-dimensional Latent Factor Models
Keyword
Benign overfitting; Classification; Deep neural nets; Dimension reduction; Discriminant analysis; Empirical risk minimization; High-dimensional statistics; Latent factor models; Minimax optimal rates of convergence; Overparametrization
Application summary
This proposal concerns the problem of classification, an important topic of statistical learning theory. High-dimensional latent factor models that have a low-dimensional, hidden structure, will be formulated to guarantee successful statistical classification performance based on suitable projections of the high-dimensional data. This project offers a principled way to analyze such high-dimensional datasets and will provide computationally efficient classification rules. The properties of our classifiers will be contrasted with others, including deep neural nets and interpolating classifiers. Results of this research are also expected to advance understanding on how to achieve optimal classification. Our proposed systematic treatment is a timely addition to the methodological and limited theoretical results in this important area. This project has important applications to recent advances in immunology and cancer studies, which revealed that hidden mechanisms can be directly connected to health outcomes. Specifically, this project constructs novel classifiers based on principal component analysis with a necessary debiasing part followed by linear discriminant analysis and develops their statistical and computational properties. In addition, this research aims to develop minimax adaptive bounds for the excess misclassification error of any classifier under general latent factor model specifications and to prove that the new methods achieve these bounds, thereby establishing their rate optimality. This project also studies of the important overparametrized classifiers that interpolate the data, but still possess good classification power, as well as the classifiers that base on deep neural nets after projection. Finally, the usefulness of the new techniques will be demonstrated via applications to data from immunology, neuroscience, economics, genomics and proteomics. This endeavor requires the close cooperation of an interdisciplinary team of researchers and graduate students at the University of Toronto. The proposed research will nurture the close ties between the PI and his scientific collaborators in Statistics and Data Sciences at Cornell University and Computational Biology and Immunology at the University of Pittsburgh. The results will be reported to the statistical, machine learning and computational biology communities at conferences, in reports made available on the world wide web, and in papers that will be submitted to leading statistical, machine learning and computational biology journals. Free software that implements the developed methodology will be made available on the web in a readily implementable form. Special effort will be made to involve undergraduates in parts of this research, as well as attracting master and Ph.D. students. Outreach to high school students and teachers will be done through the outreach programs at the Univeristy of Toronto with solid partnership with local schools, promoting equity and inclusion.