ANZIAM  J.  48 (2006), 11-22
Parallelisation of sparse grids for large scale data analysis

Jochen Garcke
  Centre for Mathematics and its Applications
  Mathematical Sciences Institute
  Australian National University
  Canberra ACT 0200
  Australia
  jochen.garcke@anu.edu.au
Markus Hegland
  Centre for Mathematics and its Applications
  Mathematical Sciences Institute
  Australian National University
  Canberra ACT 0200
  Australia
  markus.hegland@anu.edu.au
and
Ole Nielsen
  Centre for Mathematics and its Applications
  Mathematical Sciences Institute
  Australian National University
  Canberra ACT 0200
  Australia
  ole.nielsen@ga.gov.au


Abstract
Sparse grids are the basis for efficient high dimensional approximation and have recently been applied successfully to predictive modelling. They are spanned by a collection of simpler function spaces represented by regular grids. The sparse grid combination technique prescribes how approximations on a collection of anisotropic grids can be combined to approximate high dimensional functions.

In this paper we study the parallelisation of fitting data onto a sparse grid. The computation can be done entirely by fitting partial models on a collection of regular grids. This allows parallelism over the collection of grids. In addition, each of the partial grid fits can be parallelised as well, both in the assembly phase, where parallelism is done over the data, and in the solution stage using traditional parallel solvers for the resulting PDEs. Using a simple timing model we confirm that the most effective methods are obtained when both types of parallelism are used.
Download the article in PDF format (size 154 Kb)

Australian Mathematical Publishing Association Inc. ©  Australian MS