Canada Spends Logo

Explore our Budget 2026 analysis

Grant: Discovery Grants Program - Individual

McMaster University

$21,000.00

Measure-valued Stochastic Process and Applications

FY 2023-24

Summary
A fundamental issue in scientific study is to understand the rules or laws that lead to what we observe, and to apply them in forecasting and prediction. The rules or laws are usually described through mathematical models. An important class of models is the family of measure-valued processes that are used in modelling the behaviour of complex systems. Examples include the evolution of genotype frequencies of a population, the ensembles of interacting particle systems in statistical physics, and the document representations for the purpose of information retrieval. Their applications are found in many disciplines including biology, chemistry, computer science, ecology, information theory, machine learning, and physics. Recent progresses in science and technology, in particular in data science, pose a strong demand for further investigation of existing models, and the development of new models and techniques. The proposed research is in the broad area of measure-valued processes and their applications. The long term objective is to advance the theory of random measures and measure-valued processes, to develop models and algorithms for statistical inference, and to pursue applications of these results in science and engineering. The focus in the next five years will be on the following three short term objectives. The first objective is to understand the asymptotic behaviour of several Bayesian non-parametric models. These are random measures serving as priors in Bayesian statistics when the number of parameters is infinite. Results on asymptotic behaviours are important both for understanding the consistency of posteriors and for comparisons with their non-Bayesian counterparts. The second objective is to study a measure-valued process that generalizes the Fleming-Viot process in population genetics.  Given a population of individuals, one would like to know how the individual behaviour (microscopic) contributes to the population behaviour (macroscopic). The connection between the two depends on the individual interactions. This research will focus on populations where individuals either move individually or interact with others through information exchange. The goal is to understand the relation between the macroscopic behaviour and the corresponding microscopic mechanism. Distribution uncertainty is a fundamental issue in many subjects. The third objective is the study of the two-armed bandit problem with distribution uncertainty.  This is a reinforcing learning process that aims to maximize the rewards under the given uncertainties. The proposed research will introduce new criteria for asymptotically optimal solutions, and develop concrete algorithms that lead to such solutions.  The success of this research will not only advance mathematical knowledge but also lay the theoretical foundation for statistical inference in many areas. Potential applications include clinical trials, communication network, and financial engineering.
Program
Discovery Grants Program - Individual
Location
Ontario, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data

As published

Application id
RGPIN-2022-04965
Name nom
Feng, Shui
Department département
Mathematics and Statistics
Organization id
27
Institution établissement
McMaster University
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2023
Competition year année de concours
2022
Award amount
21000
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
4706
Research subject group en
Evolution and ecology
Research subject group fr
Evolutionary theory
Research subject en
Évolution et écologie
Research subject fr
Théorie de l'évolution
Application title
Measure-valued Stochastic Process and Applications
Keyword
Bayesian nonparametrics; Central limit theorems; Dirichlet process; Diversity index; Entropy; Hierarchical model; Large deviations; Measure-valued process; Nonlinear probability; Species sampling
Application summary
A fundamental issue in scientific study is to understand the rules or laws that lead to what we observe, and to apply them in forecasting and prediction. The rules or laws are usually described through mathematical models. An important class of models is the family of measure-valued processes that are used in modelling the behaviour of complex systems. Examples include the evolution of genotype frequencies of a population, the ensembles of interacting particle systems in statistical physics, and the document representations for the purpose of information retrieval. Their applications are found in many disciplines including biology, chemistry, computer science, ecology, information theory, machine learning, and physics. Recent progresses in science and technology, in particular in data science, pose a strong demand for further investigation of existing models, and the development of new models and techniques. The proposed research is in the broad area of measure-valued processes and their applications. The long term objective is to advance the theory of random measures and measure-valued processes, to develop models and algorithms for statistical inference, and to pursue applications of these results in science and engineering. The focus in the next five years will be on the following three short term objectives. The first objective is to understand the asymptotic behaviour of several Bayesian non-parametric models. These are random measures serving as priors in Bayesian statistics when the number of parameters is infinite. Results on asymptotic behaviours are important both for understanding the consistency of posteriors and for comparisons with their non-Bayesian counterparts. The second objective is to study a measure-valued process that generalizes the Fleming-Viot process in population genetics.  Given a population of individuals, one would like to know how the individual behaviour (microscopic) contributes to the population behaviour (macroscopic). The connection between the two depends on the individual interactions. This research will focus on populations where individuals either move individually or interact with others through information exchange. The goal is to understand the relation between the macroscopic behaviour and the corresponding microscopic mechanism. Distribution uncertainty is a fundamental issue in many subjects. The third objective is the study of the two-armed bandit problem with distribution uncertainty.  This is a reinforcing learning process that aims to maximize the rewards under the given uncertainties. The proposed research will introduce new criteria for asymptotically optimal solutions, and develop concrete algorithms that lead to such solutions.  The success of this research will not only advance mathematical knowledge but also lay the theoretical foundation for statistical inference in many areas. Potential applications include clinical trials, communication network, and financial engineering.