Please use this identifier to cite or link to this item: http://repositorio.unicamp.br/jspui/handle/REPOSIP/65364
Type: Artigo de periódico
Title: Efficient implementation and benchmark of interior point methods for the polynomial L-1 fitting problem
Author: Oliveira, ARL
Nascimento, MA
Lyra, C
Abstract: Interior point methods specialized to the L-1 fitting problem are surveyed and the affine-scaling primal method is presented. Their main features are highlighted and improvements are proposed for polynomial fitting problems. For such problems, a careful handling of data avoids storing of matrices for the interior point approaches. Moreover, the computational complexity of iterations is reduced. An inexpensive way to compute a basic solution, using interpolation, is also provided. Extensive numerical experiments are carried out, including comparisons with a specialized simplex method. In general, the interior point methods performed better than the simplex approach. Among the interior point methods investigated, the dual affine scaling version was the most efficient. (C) 2000 Elsevier Science B.V. All rights reserved.
Subject: L-1 curve fitting
L-1 regression problems
L-1-based statistical analysis
interior point methods
linear programming
data mining
Country: Holanda
Editor: Elsevier Science Bv
Rights: fechado
Identifier DOI: 10.1016/S0167-9473(00)00006-2
Date Issue: 2000
Appears in Collections:Unicamp - Artigos e Outros Documentos

Files in This Item:
File Description SizeFormat 
WOS000166220500001.pdf381.52 kBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.