一维搜索试探：斐波那契数列法的MATLAB实现

适用于（凸）优化理论入门学习者。

一、基本思想

• 一维搜索：目标函数为一元单值函数 的优化问题的迭代求解方法
• 试探法：取试探点和函数值做比较，使得包含极小点的搜索区间不断缩短，适合单峰函数。
• 斐波那契数列法的要点：

        1.区间长度的缩短率，由斐波那契数确定 : 

           ,   .

        2.迭代次数 n 需要事先给定，n可由初始区间长度和精度要求计算：

二、斐波那契数列法的MATLAB实现

•  主函数：

function [result] = fibonacci_array(A,B)
fib=[1,1,2,3,5,8,13,21,34];
n=4;
%n=6;
a = A;
b = B;

k=1;
lambda=a+(fib(n-k)/fib(n-k+2))*(b-a);
mu=a+(fib(n-k+1)/fib(n-k+2))*(b-a);

display=[k,a,b,lambda,mu,diox(lambda),diox(mu)];
disp(display);
delta = 0.05;
%delta = 0.01;

while (k<= n)
        if (diox(lambda)>diox(mu))
            a = lambda;
            lambda = mu;
            mu=a+(fib(n-(k+1)+1)/fib(n-(k+1)+2))*(b-a);
            
            if (k == n-2)
                display=[k+1,a,b,lambda,mu,diox(lambda),diox(mu)];
                disp(display);
                
                lambda1 = lambda;
                mu1 = mu+ delta;
                if diox(lambda1)>diox(mu1)
                    k = k+2;
                    a1 = lambda1;
                    b1 = b;
                else
                    k = k+2;
                    a1 = a;
                    b1 = lambda1;
                end
                display=[k,a1,b1,lambda1,mu,diox(lambda1),diox(mu)];
                result = (a1+b1)/2;
                disp(display);
                break;
            else
                k = k+1;
                display=[k,a,b,lambda,mu,diox(lambda),diox(mu)];
                disp(display);
            end
            
         else
            b = mu;
            mu = lambda;
            %
            if (n-(k+1))<1
                break;
            end
            lambda = a+(fib(n-(k+1))/fib(n-(k+1)+2))*(b-a);
            if (k == n-2)
                display=[k+1,a,b,lambda,mu,diox(lambda),diox(mu)];
                disp(display);
                
                lambda1 = lambda;
                mu1 = mu+ delta;
                if diox(lambda1)>diox(mu1)
                    k = k+2;
                    a1 = lambda1;
                    b1 = b;
                else
                    k = k+2;
                    a1 = a;
                    b1 = lambda1;
                end
                display=[k,a1,b1,lambda1,mu,diox(lambda1),diox(mu)];
                disp(display);
                result = (a1+b1)/2;
                break;
            else
                k = k+1;
                display=[k,a,b,lambda,mu,diox(lambda),diox(mu)];
                disp(display);
            end
        end

end

end

• 目标函数：

  function [FX] = diox(X)
  FX=X^2+4*cos(X);
  end