Tensor Approaches for Simulating Kinetic Systems

Abstract

We propose a comprehensive collaborative ef fort to tackle the wide-ranging theoretical and computational challenges associated with approximating time-dependent high-dimensional PDEs on tensor manifolds. The core of our proposal centers around advancing tensor- based algorithms that uphold the intrinsic structures within a reduced low-rank representation of dynamics (Thrust 1), as well as to push the computing capabilities of extremely high-dimensional equations via data-driven learning approaches for tensor representations and model reduction (Thrust 2). The efficiency and robustness of tensor-based algorithms developed in Thrusts 1 and 2 hinges on the comprehensive development of tensor decomposition and optimization algorithms (Thrust 3).

Document Details

Document Type
DoD Grant Award
Publication Date
Feb 06, 2025
Source ID
FA95502410254

Entities

People

  • Jingmei Qiu

Organizations

  • Air Force Office of Scientific Research
  • United States Air Force
  • University of Delaware

Tags

Fields of Study

  • Computer science

Readers

  • Linear Algebra
  • Neural Network Machine Learning.