Variable metric methods for a class of extended conic functions. (English) Zbl 0548.90062

Lukšan, Ladislav

Kybernetika 21, 96-107 (1985).

The paper contains a description and an analysis of two variable metric algorithms for unconstrained minimization which find a minimum of an extended conic function after a finite number of steps provided it is possible to compute the derivatives of the model function at an arbitrary point \(x\in R_ n\). Moreover, the developed theory is applied to a special class of the exponential type of extended conic functions.

MSC:

90C30	Nonlinear programming
49M37	Numerical methods based on nonlinear programming
65K05	Numerical mathematical programming methods

Keywords:

finite step convergence; variable metric algorithms; unconstrained minimization; extended conic functions

References:

[1]	P. Bjørstad, J. Nocedal: Analysis of a new algorithm for one-dimensional minimization. Computing 22 (1979), 2, 93-100.
[2]	C. G. Broyden: Quasi-Newton methods and their application to function minimization. Math. Comp. 21 (1967), 99, 368-381. · Zbl 0155.46704 · doi:10.2307/2003239
[3]	W. C. Davidon: Conic approximations and collinear scalings for optimizers. SIAM J. Numer. Anal. 77 (1980), 2, 268-281. · Zbl 0424.65026 · doi:10.1137/0717023
[4]	L. Lukšan: Quasi-Newton methods without projections for unconstrained minimization. Kybernetika 18 (1982), 4, 290-306. · Zbl 0514.65047
[5]	L. Lukšan: Conjugate gradient algorithms for conic functions. Computing · Zbl 0622.65045
[6]	D. C. Sorensen: The Q-superlinear convergence of a collinear scaling algorithm for unconstrained optimization. SIAM J. Numer. Anal. 77 (1980), I, 84-114. · Zbl 0428.65040 · doi:10.1137/0717011