% 算法步骤如下：

% 1.设定种群个体数和需要迭代的次数。

% 2.选择父代中的个体按照公式:

% z1=sqrt(-2*ln(u1))*sin(2*pi*u2)*m，

% z2=sqrt(-2*ln(u1))*cos(2*pi*u2)*m

% 这里u1,u2都是随机值，m是控制因子，演化次数越多m，

% m越小，父代通过与z1，z2相加得到后代。

% 3.计算后代的适应性。

% 4.选择后代中最优的适应性作为全局最优适应性。

%% matlab 代码开始

clc;

clear all;

close all;

%% 绘出所求函数图形

[x,y] = meshgrid(-100:100,-100:100);

sigma = 50;

img = (1/(2*pi*sigma^2))*exp(-(x.^2 + y.^2)/(2*sigma^2));

mesh(img);

axis([-100,100,-100,100,2e-5,8e-5]);

hold on

%% 初始化种群

pop_size = 20;%种群规模

pop = struct([]);

iterative = 100;

for i = 1:pop_size

pop(i).x = -100 + 200*rand();

pop(i).y = -100 + 200*rand();

pop(i).fit = compute_fit(pop(i));

end

pop_best = pop(1);

%% 演化开始

p1 = plot3(pop_best.x +100,pop_best.y+100,pop_best.fit,‘k.‘,‘markersize‘,30);

for i = 1:iterative

set(p1, ‘xdata‘,pop_best.x+100,‘ydata‘,pop_best.y+100,‘zdata‘,pop_best.fit)

drawnow

pause(.1)

[pop,pop_best] = select_and_combin(pop,pop_best,pop_size,i,iterative);

end

function [new_pop,pop_best] = select_and_combin(pop,pop_best,pop_size,k,iterative)

mul = (iterative - k) / iterative;

new_pop = pop;

% z1=sqrt(-2*ln(u1))*sin(2*pi*u2)*m，

% z2=sqrt(-2*ln(u1))*cos(2*pi*u2)*m

for i = 1:pop_size

%随机数

u1 = rand();

u2 = rand();

z1 = sqrt(-2*log(u1))*sin(2*pi*u2)*mul;

z2 = sqrt(-2*log(u1))*cos(2*pi*u2)*mul;

%变异

new_pop(i).x = pop(i).x + z1;

new_pop(i).y = pop(i).y + z2;

%计算适应值

new_pop(i).fit = compute_fit(new_pop(i));

if new_pop(i).fit < pop(i).fit

new_pop(i) = pop(i);

end

if pop_best.fit < new_pop(i).fit

pop_best = pop(i);

end

end

end

function re = compute_fit(pop)

x = pop.x;

y = pop.y;

sigma = 50;

if x < -100 || x > 100 || y < -100 || y > 100

re = 0;

else

re =(1/(2*pi*sigma^2))*exp(-(x^2 + y ^2)/(2 * sigma^2));

end

http://www.cnblogs.com/tiandsp/p/3165388.html

原文：http://www.cnblogs.com/Kermit-Li/p/4051915.html