Overcoming the Curse of Dimension: Methods Based on Sparse Representation and Adaptive Sampling

Abstract

A major issue in modeling and computation is how to handle high dimensional problems. We can divide these high dimensional problems into two classes: Moderately high dimensional problems or very high dimensional problems. In the former class, we have problems such as the Boltzmann equation, whose dimensionality is high but they are still amendable to grid-based methods. In the latter class we have problems such as exploration of the configuration space of a large molecule. These problems often involve hundreds of thousands of dimensions, and methods based on fixed grids are far from being adequate. We have explored various ways of handling these problems using the sparse representation or the adaptive sampling.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Feb 28, 2011
Accession Number
ADA564054

Entities

People

  • E. Weinan

Organizations

  • Princeton University

Tags

Communities of Interest

  • Energy and Power Technologies

DTIC Thesaurus Topics

  • Boltzmann Equation
  • Chemical Engineering
  • Chemistry
  • Collecting Methods
  • Computational Fluid Dynamics
  • Computational Science
  • Density Functional Theory
  • Differential Equations
  • Equations
  • Experimental Data
  • Fluid Dynamics
  • Fokker Planck Equations
  • Mathematics
  • Monte Carlo Method
  • Multiscale Modeling
  • Partial Differential Equations
  • Sampling

Fields of Study

  • Computer science

Readers

  • Calculus or Mathematical Analysis
  • Distributed Systems and Data Platform Development
  • Systems Analysis and Design

Technology Areas

  • Space