MATLAB Program to Find A Function Minimum

Using a Random Search Method

by Namir Shammas

The following program calculates the minimum point of a multi-variable function using random search method. The search selects random points within a given range for each variable.

The function Random_Search1 has the following input parameters:

N - number of variables
XLo - array of lower values
XHi - array of higher values
Eps_Fx - tolerance for difference in successive function values
MaxIter - maximum number of iterations
myFx - name of the optimized function

The function generates the following output:

X - array of optimized variables
BestF - function value at optimum
Iters - number of iterations

Here is a sample session to find the optimum for the following function:

y = 10 + (X(1) - 2)^2 + (X(2) + 5)^2

The above function resides in file fx1.m. The search for the optimum 2 variables has the search range between [-10 -10] and [10 10]. The search employs a maximum of 10000 iterations and a function tolerance of 1e-7:

>> [XBest,BestF,Iters]=Random_Search1(2, [-10 -10], [10 10], 1e-7, 10000, 'fx1')

XBest =

2.4200 -4.6230

BestF =

10.3185

Iters =

10000

Keep in mind that your results will most likely be different since the algorithm uses random numbers.

Here is the MATLAB listing:

function y=fx1(X, N)
  y = 10 + (X(1) - 2)^2 + (X(2) + 5)^2;
end

function [XBest,BestF,Iters]=Random_Search1(N,XLo,XHi,Eps_Fx,MaxIter,myFx)
% Function performs multivariate optimization using the
% random search method.
%
% Input
%
% N - number of variables
% XLo - array of lower values
% XHi - arra of higher values
% Eps_Fx - tolerance for difference in successive function values
% MaxIter - maximum number of iterations
% myFx - name of the optimized function
%
% Output
%
% XBest - array of optimized variables
% BestF - function value at optimum
% Iters - number of iterations
%

XBest = XLo + (XHi - XLo) * rand(N);
BestF = feval(myFx, XBest, N);;
LastBestF = 100 * BestF + 100;

Iters = 0;

for i=1:MaxIter
  Iters = Iters + 1;
  X = XLo + (XHi - XLo) * rand(N);
  F = feval(myFx, X, N);
  if F < BestF
    XBest = X;
    LastBestF = BestF;
    BestF = F;
  end

  if abs(LastBestF - BestF) < Eps_Fx
    break
  end
end

BACK