牛顿法在回归分析中的应用

      前面已经讲过一些参数优化的方法，现在来讲牛顿法在回归分析中的应用。

下面给出实验结果

main.m

x = load('ex3x.dat');      
y = load('ex3y.dat');      
      
trustRegionBound = 1000;      
x = [ones(size(x,1),1) x];      
meanx = mean(x);%求均值      
sigmax = std(x);%求标准偏差      
x(:,2) = (x(:,2)-meanx(2))./sigmax(2);      
x(:,3) = (x(:,3)-meanx(3))./sigmax(3);      
itera_num = 2000; %尝试的迭代次数      
sample_num = size(x,1); %训练样本的次数      
jj=0.00001;  
figure      
alpha = [0.1];%因为差不多是选取每个3倍的学习率来测试，所以直接枚举出来      
plotstyle = {'b-'}; 
theta_grad_descent = zeros(size(x(1,:)));    
%% 牛顿法
theta = zeros(size(x,2),1); %theta的初始值赋值为0      
Jtheta = zeros(itera_num, 1);
Q=x'*x;
for i = 1:itera_num %计算出某个学习速率alpha下迭代itera_num次数后的参数             
        Jtheta(i) = (1/(2*sample_num)).*(x*theta-y)'*(x*theta-y);%Jtheta是个行向量      
        grad = (1/sample_num).*x'*(x*theta-y);          
        d=-(inv(Q)*grad);    
        a=(grad'*grad)/(grad'*Q*grad);    
        theta = theta + a*d;       
 end      
K(1)=Jtheta(500)      
plot(0:999, Jtheta(1:1000),'b--','LineWidth', 4);    
hold on 

%%  
legend('Newton method');      
xlabel('Number of iterations')      
ylabel('Cost function')