vix.ing · top · new · best · stats

The Differentiation of Pseudo-Inverses and Nonlinear Least Squares Problems Whose Variables Separate

1973/04/01 by Gene H. Golub, G. H. Golub, V. Pereyra +1 · 1,427 citations
Computer Science · Mathematics · #Advanced Optimization Algorithms Research #Combinatorics #Geometry #Inverse #Iterative Methods for Nonlinear Equations #Mathematics #Matrix (chemical analysis) #Matrix Theory and Algorithms #Rank (graph theory)

paper · doi:10.1137/0710036

published in SIAM Journal on Numerical Analysis 10(2), 413-432 (Society for Industrial and Applied Mathematics)

openalex publication_date 1973/04/01 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/25

Abstract

For given data (ti ,yi ),i = 1, ⋯ ,m, we consider the least squares fit of nonlinear models of the form η (\bf a,\boldsymbol α ;t) = ∑ j = 1n aj φ j (\boldsymbol α ;t), \bf a ∈ Rn , \boldsymbol α ∈ Rk . For this purpose we study the minimization of the nonlinear functional r(\bf a,\boldsymbol α ) = ∑i = 1m ( yi - η ( \bf a,\boldsymbol α ,ti ) )2 . It is shown that by defining the matrix \ \bf Φ (\boldsymbol α )\ i,j = φ j (\boldsymbol α ;ti ), and the modified functional r2 (\boldsymbol α ) = ‖ \bf y - \bf Φ (\boldsymbol α )\bf Φ ^ + (\boldsymbol α )\bf y ‖22 , it is possible to optimize first with respect to the parameters \boldsymbol α , and then to obtain, a posteriors, the optimal parameters \bf a. The matrix \bf Φ ^ + (\boldsymbolα ) is the Moore–Penrose generalized inverse of \bf Φ (\boldsymbolα ). We develop formulas for the Frechet derivative of orthogonal projectors associated with \bf Φ (\boldsymbolα ) and also for \bf Φ ^ + (\boldsymbolα ), under the hypothesis that \bf Φ (\boldsymbolα ) is of constant (though not necessarily full) rank. Detailed algorithms are presented which make extensive use of well-known reliable linear least squares techniques, and numerical results and comparisons are given. These results are generalizations of those of H. D. Scolnik [20] and Guttman, Pereyra and Scolnik [9].

Cited by

Related