Local optimization methods and software

Method
Author(s)
Use of gradient?
Fitness function
Software
SIMPLEX
Nelder & Mead
 NO ANY NR & IDL: amoeba
MINUIT: simplex
POWELL
Powell
 NO ANY NR & IDL: powell
STEPPEST DESCENDENT
YES ANY?
CONGUGATE GRADIENT
Fletcher-Reeves
Polack-Ribiere
 YES(NO) ANY NR: frprmn
VARIABLE METRIC
Davidon-Fletcher-Powell
YES ANY MINUIT: migrad
VARIABLE METRIC
Broyden-Fletcher-
Goldfarb-Shanno
YES ANY NR & IDL: dfpmin
LM (Unbounded)
Levenberg-Marquardt
 YES? ONLY CHI-SQUARE IDL: lmfit
NR: mrqmin
LM (Bounded)
Levenberg-Marquardt
 YES? ONLY CHI-SQUARE IDL: mpfit
MINPACK
GENERALIZED
REDUCED
GRADIENT
Windward Tech.
 ANY
(with constraints)		 IDL: constrained-min
DHC
D. Yuret
 NO ANY IDL, C and Fortran


Notes:
Software: NR ( Numerical Recipes book), IDL ( Interactive data language), MINUIT from CERN, MINPACK ( single precision sminpack, double precision minpack) from Netlib.




[Back to opt-L]



Last modified: 6 August 1999
Christoph Michetschlaeger and M. Sanchez del Rio (srio@esrf.fr)
~