Local minimization of smooth functions without evaluating derivatives

Experimental files for Matlab or octave - free and without any guarantee. Please report errors to jarre@hhu.de
Optimization and Engineering, 2016, 1-28.
Calibration by Optimization Without Using Derivatives

The paper describes an application of min_f.m (see below) and a summary of the concepts used in min_f.m
(The paper refers to Version0 of March 2015, where the case n=1 and n>1 were treated in two separate m-files, now merged to the single file min_f.m)

min_f.m -- a single m-file for n-dimensional local minimization (for n>= 1) - unconstrained or subject to simple bounds (last change: Feb. 9, 2018)

Test function of 1 variable with unique minimizer in [0,1]:

Test function of 1 variable for min_f

Test functions of n variables with unique minimizers:

Chained Rosenbrock function

Modified Stybilsky-Tang-function

Some exponential sum

Preprint, May 2017:
A Derivative-Free and Ready-to-Use NLP Solver for Matlab or Octave

The paper describes the calling routine for min_fc.m (see below) and a summary of the concepts used in min_fc.m

min_fc.m -- a single m-file for nonlinear minimization subject to linear and nonlinear equality and inequality constraints. (No subroutines for any derivatives needed.) May 5. 2017

Sample file how to define objective function and constraints and how to call min_fc: :

Calling routine for min_fc

A sample objective function and sample constraints function:

Sample objective function

Sample constraint function

m-files of previous versions

mehrotramsocp.m -- a single m-file for an interior point method for solving mixed second order cone programs (MSOCP)

  • Verantwortlich für den Inhalt: Lehrstuhl Mathematische Optimierung
  • E-Mail: jarre@hhu.de