粒子群优化算法(PSO)找最优解

关于粒子群优化算法(PSO)的解释:http://blog.csdn.net/myarrow/article/details/51507671 (等网上各种参考资料)

用PSO找函数极值。

function [ f ] = fun( x, y )
%UNTITLED2 此处显示有关此函数的摘要
%   此处显示详细说明
a = 50;
b = 50;

f = (x-a)^2 + (y-b)^2;

end

构造了一个二元函数,很明显,有极小值,在(50, 50)。

粒子群优化算法(PSO)找最优解_第1张图片

用五个粒子去找。每个点对应的函数输出值即视为粒子的适应度。

 1 clear all;
 2 clc;
 3 % initialize
 4 n = 5;
 5 P = rand(5, 2) * 100;
 6 p = cell(5, 1);
 7 for i = 1: n
 8     p(i, 1) = {[P(i, 1), P(i, 2)]};
 9 end
10 v = rand(5, 2) * 0.5;
11 vmax = 0.5;
12 pbest = p;
13 gbest = cell(1);
14 fit_v = zeros(5, 1);
15 key = 1;
16 for i = 1: n
17     fit_v(i, 1) = fun(p{i}(1), p{i}(2));
18     if( fit_v(i, 1) < fit_v(key, 1) )
19         key = i;
20     end
21 end
22 gbest(1) = pbest(key, 1);
23 % loop
24 t = 1;
25 while(t < 300)
26     for i = 1: n
27         for j = 1: 2
28             v(i, j) = v(i, j) + 2*rand()*(pbest{i, t}(j)-p{i}(j)) + 2*rand()*(gbest{t}(j)-p{i}(j));
29             if( abs(v(i, j)) > vmax )
30                 v(i, j) = abs(v(i, j)) / v(i, j) * vmax;
31             end
32             p{i}(j) = p{i}(j) + v(i, j);
33         end
34         fit_v_temp = fun(p{i}(1), p{i}(2));
35         if( fit_v_temp < fit_v(i, t) )
36             fit_v(i, t+1) = fit_v_temp;
37             pbest(i, t+1) = p(i);
38         else
39             fit_v(i, t+1) = fit_v(i, t);
40             pbest(i, t+1) = pbest(i, t);
41         end
42     end
43     t = t + 1;
44     key = 1;
45     for i = 1: n
46         if( fit_v(i, t) < fit_v(key, t) )
47             key = i;
48         end
49     end
50     gbest(t) = pbest(key, t);
51 end

运行结果:[50.0030796073836,50.0004541255463]

粒子群优化算法(PSO)找最优解_第2张图片 

      图1 gbest轨迹(红点表示最后收敛位置)

粒子群优化算法(PSO)找最优解_第3张图片

          图2 各个粒子的适应度曲线

 

可以发现在100次左右就基本收敛了。

 

转载于:https://www.cnblogs.com/Zzz-y/p/7978090.html

你可能感兴趣的:(粒子群优化算法(PSO)找最优解)