ossrs

A random search method[1] for the optimization of a function of n variables.

Description

Optimized Step Size Random Search(OSSRS) A random search method[1] for the optimization of a function of n variables. Can be used as a start-up before using a determistic method like Simplex method of Nelder and Mead to arrive at the region of the optimum and then switch to a deterministic search.
 
The function to be minimized is to be deifned in the m function called fun.m. 'n' is the number of variables. x is the vector of initial variables. 'std' can be set based on the range of variation of the variables- about 10% of x- the initial values-should be right.
  
The following demo function is included:
  
     Minimize: f(x)=(x(1)-5)^2+(x(2)+8)^2
     Solution: [5,-8]
  
Reference:
 
Sheela Belur V, An Optimized Step Size Random Search , Computer Methods in Appl. Mech & Engg, Vol. 19, 99-106 , 1979
 
Tested under MATLAB 4.2c & 5

• MATLAB release: MATLAB 5.2 (R10)

[Antoni Vaello]

 

[Victor Zakharov]

 The algorith was tested by my. It good work
for quadratic function and near them.
  In code there are mistake with calculation
 summary function evaluation. I may to propouse the test results.==Victor

[Andrew Daubenspeck]

Seems to work well with the infamous Rosenbrock function. Documentation could be better.