Some Topics in Large Deviations Theory for Stochastic Dynamical Systems Public Deposited

Downloadable Content

Download PDF
Last Modified
  • March 21, 2019
  • Chen, Jiang
    • Affiliation: College of Arts and Sciences, Department of Statistics and Operations Research
  • In this dissertation, we study large deviations problems for stochastic dynamical systems. First, we consider a family of Stochastic Partial Differential Equations (SPDE) driven by a Poisson Random Measure (PRM) that are motivated by problems of chemical/pollutant dispersal. We established a Large Deviation Principle (LDP) for the long time profile of the chemical concentration using techniques based on variational representations for nonnegative functionals of general PRM. Second, we develop a LDP for small Poisson noise perturbations of a general class of deterministic infinite dimensional models. SPDEs driven by PRM have been proposed as models for many different physical systems. The approach taken here, which is based on variational representations, reduces the proof of the LDP to establishing basic qualitative properties for controlled analogues of the underlying stochastic system. Third, we study stochastic systems with two time scales. Such multiscale systems arise in many applications in engineering, operations research and biological and physical sciences. The models considered in this dissertation are usually referred to as systems with "full dependence", which refers to the feature that the coefficients of both the slow and the fast processes depend on both variables. We establish a LDP for such systems with degenerate diffusion coefficients. The last part of this dissertation focuses on numerical schemes for computing invariant measures of reflected diffusions. Reflected diffusions in polyhedral domains are commonly used as approximate models for stochastic processing networks in heavy traffic. Stationary distributions of such models give useful information on the steady state performance of the corresponding stochastic networks and thus it is important to develop reliable and efficient algorithms for numerical computation of such distributions. We propose and analyze a Monte-Carlo scheme based on an Euler type discretization. We prove an almost sure consistency of the appropriately weighted empirical measures constructed from the simulated discretized reflected diffusion to the true diffusion model. Rates of convergence are also obtained for certain class of test functions. Some numerical examples are presented to illustrate the applicability of this approach.
Date of publication
Resource type
Rights statement
  • In Copyright
  • Budhiraja, Amarjit
  • Doctor of Philosophy
Degree granting institution
  • University of North Carolina at Chapel Hill
Graduation year
  • 2013

This work has no parents.