梯度下降Gradient Descent matlab实现

梯度下降Gradient Descent matlab实现

文章参考：http://www.cnblogs.com/pinard/p/5970503.html

代码：

function theta = gradientDescent(x, y, a, acc)
% a 步长  acc 损失函数可满足的最小值
[h,w]=size(x);
theta=zeros(w+1,1);
temp=theta;
c=ones(h,1);
x=[c x];

J=Inf;
while(J>acc)
    for j=1:w+1
        sum=0; 
        for k=1:h
            sum=sum+(x(k,:)*theta-y(k))*x(k,j);
        end
        %sum %
        temp(j)=theta(j)-a*(sum)/h;
    end
    for j=1:w+1    
        theta(j)=temp(j);
    end
    J=0;
    for i=1:h   
        J=J+(x(i,:)*theta-y(i))^2;
    end
    J=J/(2*h);
end

数据：