Abstract
The traditional method of data fitting is by the least squares (l 2) technique. When the data is good—reasonably accurate with normally distributed errors—this method is ideal. When the data is bad—contaminated by occasional wild values—then the l 1 technique (minimizing sums of absolute values of residuals) has much to recommend it. This paper surveys the strategy of a globally and superlinearly convergent algorithm to minimize sums of absolute values of C 2 functions. The approach to be presented is closely related to the use of a certain, piecewise differentiable penalty function to solve nonlinear programming problems.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
Bibliography
Bartels, R.H. and Conn, A.R. Linearly constrained discrete l 1 problems. ACM Trans. on Math. Software 6 (1980) 594–608.
Coleman, T.F. and Conn, A.R. Second-order conditions for an exact penalty function. Math. Prog. 19 (1980) 178–185.
Coleman, T.F. and Conn, A.R. Nonlinear programming via an exact penalty function: global analysis. Math. Prog. to appear.
Colman, T.F. and Conn, A.R. Nonlinear programming via an exact penalty function: asymptotic analysis. Math. Prog. to appear.
El-Attar, R.A.; Vidyasagar, M. and Dutta, S.R.K. An Algorithm for l 1-norm minimization with application to nonlinear l 1-approximation. SIAM J. Num. Anal. 16 (1979) 70–86.
Overton, M.L. and Murray, W. A Projected Lagrangian algorithm for nonlinear l 1 optimization. SIAM J. Sci. and Stat. Comp. to appear.
Author information
Authors and Affiliations
Editor information
Rights and permissions
Copyright information
© 1982 Springer-Verlag
About this paper
Cite this paper
Bartels, R.H., Conn, A.R. (1982). An approach to nonlinear l 1 data fitting. In: Hennart, J.P. (eds) Numerical Analysis. Lecture Notes in Mathematics, vol 909. Springer, Berlin, Heidelberg. https://doi.org/10.1007/BFb0092959
Download citation
DOI: https://doi.org/10.1007/BFb0092959
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-11193-1
Online ISBN: 978-3-540-38986-6
eBook Packages: Springer Book Archive