Variable selection and control in least squares problems

Research Report MRR98-047

Variable selection and control in least squares problems

M.R. Osborne

Abstract: The classical technique of stepwise regression provides a paradigm for variable selection in the linear least squares problem. Trust region methods which control the size of the correctionto the current solution estimate prove attractive for nonlinear least squares problems because if their good global convergence behaviour. Recently there has been a convergence of these techniques with the realisation that the l₁ trust rtegion method also provides a form of variable selection. These results are reviewed here, and computational methods discussed.

Download paper:	PDF file (204K)
	gzipped DVI file (23K)

Download cover sheet:	PDF file (44K)
	DVI file (2K)

This service is maintained by the Mathematical Sciences Institute (MSI)
Comments to webmaster@maths.anu.edu.au URL: http://wwwmaths.anu.edu.au/