ANU Data Mining Group

One of our research projects is the development of a predictive modelling framework using sparse grids (PMSG). PMSG is a collection of Python modules which allows building of regression models of datasets and subsequent evaluation of them on new data. PMSG is based on 'sparse grids' which is an approximative representation of high dimensional spaces. PMSG are research codes which are under continual development for UNIX platforms only and they are not a final out-of-the-box product which will run anywhere for any problem.

The theory of sparse grids and PMSG have been presented in the following publications.

• Additive Sparse Grid Fitting
  Markus Hegland
  Submitted to Fifth International Conference on Curves and Surfaces organised by AFA-SMAI, Saint-Malo, France, 27 June - 3 July, 2002.
  Paper (pdf, 136 KB)

• Adaptive Sparse Grids
  Markus Hegland
  Submitted to the CTAC-2001 Conference, Brisbane, 16-18 July 2001.
  Paper (ps.gz, 142 KB)

Archives (contain all files compressed):

• Version 1.3: pm_1_3.tgz (1.1 MB)

• Version 1.2: pm_1_2.tgz (1.1 MB)

• Version 1.1: pm_1_1.tgz (11 MB) (includes data files)

• README
  Gives information on how to install and use PMSG.

• DOC
  Gives and overview and some documentation on PMSG.

• LICENSE
  A copy of the GNU General Public License.

Please contact us if you have questions, problems or ideas. All feedback is much appreciated.

Authors

Ole Nielsen (Ole.Nielsen@anu.edu.au)
Markus Hegland (Markus.Hegland@anu.edu.au)