Novel Numerical Methods for Nonlinear Stochastic PDEs and High Dimensional Computation

University Of Tennessee Knoxville, Knoxville TN

Investigators

Abstract

Many scientific, engineering, and industrial applications involve random effects. Incorporating uncertainties into mathematical models becomes indispensable in order to develop more accurate and robust mathematical models from biological, engineering, and physical applications, which makes it necessary to consider stochastic partial differential equations as they are the most commonly encountered stochastic models from applications. To seek solutions of those equations require accurate, efficient, and robust computational methods and algorithms, the demand for such methods and algorithms has never been greater. The current approaches for solving stochastic partial differential equations face considerable challenges at large scales, in the meantime, such stochastic models become more complicated as more difficult nonlinearity and noise are considered. The existing numerical approaches are not efficient to solve those problems, which in turn calls for new ideas and approaches. With rapid developments in nontraditional applied sciences such as mathematical finance, image processing, economics, and data science, there is an ever-increasing demand for efficient numerical methods for solving challenging high-dimensional problems such as computing integration and solving partial differential equations. The traditional grid-based methods are hampered by the infamous curse of dimensionality in which the amount of required computations for solving a problem grows exponentially in the dimension. The primary goal of this research project is to develop novel and efficient numerical methods to address those challenges. The project consists of two integral parts. Part I focuses on developing and analyzing efficient numerical methods for solving several nonlinear stochastic partial differential equations which arise from various scientific and engineering applications such as materials science, fluid and quantum mechanics, and optimal control. Part II will be devoted to developing novel numerical methods for high-dimensional computation with a focus on problems of high-dimensional numerical integration and high-dimensional partial differential equations. The overreaching vision of this project is to develop a framework for constructing and analyzing numerical methods for general nonlinear stochastic partial differential equations and to develop new approaches and enabling methods for overcoming the curse of dimensionality challenge for high-dimensional computation. The project will include training of graduate students. This research project develops advanced numerical methods for nonlinear stochastic partial differential equations and high-dimensional computation. Such a timely and advanced project is of great interest to the STEM community as the anticipated numerical methods and algorithms will provide much-needed enabling tools for tackling challenging problems described mathematically by stochastic partial differential equations or involved with high-dimensional computation from many scientific, engineering, and industrial applications as well as AI, machine learning, and data science. This research project is also expected to have a lasting impact on the advancement of numerical stochastic partial differential equations and high-dimensional computation. Moreover, the project will provide a valuable opportunity and resource to train Ph.D. graduate students and to help them to develop necessary applied and computational mathematics as well as AI, machine learning, and data science knowledge and skills so that they can pursue successful careers in either academia or industry in the near future. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.