Projection pursuit optimisation using a genetic algorithm

Goodman, Steve and Hunter, Andrew (1999) Projection pursuit optimisation using a genetic algorithm. Project Report. University of Sunderland, Sunderland. (Unpublished)

Documents
Projection_Pursuit_Optimisation_using_a_Genetic_Algorithm.pdf
[img]
[Download]
Request a copy
[img] PDF
Projection_Pursuit_Optimisation_using_a_Genetic_Algorithm.pdf - Whole Document
Restricted to Repository staff only

1416Kb

Abstract

This paper describes the application of a genetic algorithm to the optimisation of a data projection method known as projection pursuit. Exploratory projection pursuit, named by Friedman and Tukey (1974) is a statistical method for viewing low (typically one or two) dimensional projections of higher dimensional data. This has been applied in data analysis as a visualisation technique for examination of possible non-linear structure perhaps not reveewaled by other techniques such as Principal Component Analysis (Huber, 1985; Jones and Sibson, 1987). It is claimed that bec ause it works in low dimensional space, it can overcome the problem known as the 'curse of dimensionality' (Bellman 1961) which plagues other multivariate techniques. This is usually caused by having a high dimensional data set which contains too few observations to adequately represent the data space, a problem frequently found in real, industrial problems.

Item Type:Paper or Report (Project Report)
Additional Information:This paper describes the application of a genetic algorithm to the optimisation of a data projection method known as projection pursuit. Exploratory projection pursuit, named by Friedman and Tukey (1974) is a statistical method for viewing low (typically one or two) dimensional projections of higher dimensional data. This has been applied in data analysis as a visualisation technique for examination of possible non-linear structure perhaps not reveewaled by other techniques such as Principal Component Analysis (Huber, 1985; Jones and Sibson, 1987). It is claimed that bec ause it works in low dimensional space, it can overcome the problem known as the 'curse of dimensionality' (Bellman 1961) which plagues other multivariate techniques. This is usually caused by having a high dimensional data set which contains too few observations to adequately represent the data space, a problem frequently found in real, industrial problems.
Keywords:Genetic algorithms, pursuit optimisation
Subjects:G Mathematical and Computer Sciences > G400 Computer Science
Divisions:College of Science > School of Computer Science
ID Code:3386
Deposited By: Tammie Farley
Deposited On:25 Sep 2010 19:43
Last Modified:13 Mar 2013 08:47

Repository Staff Only: item control page