神经网络作业: NN LEARNING Coursera Machine Learning（Andrew Ng） WEEK 5

在WEEK 5中，作业要求完成通过神经网络(NN)实现多分类的逻辑回归(MULTI-CLASS LOGISTIC REGRESSION)的监督学习(SUOERVISED LEARNING)来识别阿拉伯数字。作业主要目的是感受如何在NN中求代价函数（COST FUNCTION）和其假设函数中各个参量(THETA)的求导值（GRADIENT DERIVATIVE)（利用BACKPROPAGGATION）。

难度不高，但问题是你要习惯使用MATLAB的矩阵QAQ，作为一名蒟蒻，我已经狗带了。以下代核心部分的代码希望给被作业卡住的同学一些帮助。但请不要照搬代码哦~不要~不要~

 ty = zeros(m, num_labels);

 for i=:m
     for j=:num_labels
         if y(i)==j
             ty(i,j) = ;
         end
     end
 end

 a1 = X;
 a1 = [ones(size(a1,),) a1];
 z2 = a1 * Theta1';
 a2 = sigmoid(z2);
 a2 = [ones(size(a2,),) a2];
 z3 = a2 * Theta2';
 a3 = sigmoid(z3);

 for i=:m
     for j=:num_labels
         J = J - log(-a3(i,j))*(-ty(i,j))/m-log(a3(i,j))*ty(i,j)/m;
     end
 end

 %size(J,)
 %size(J,)

 d3 = a3 - ty;
 d2 = (d3 * Theta2(:,:end)).*sigmoidGradient(z2);
 Theta1_grad = Theta1_grad + d2'*a1/m;
 Theta2_grad = Theta2_grad + d3'*a2/m;

 % -------------------------------------------------------------
 JJ=;

  for i=:size(Theta1,)
     for j=:size(Theta1,)
                 JJ = JJ + Theta1(i,j)*Theta1(i,j)*lambda/(m*);
     end
  end
  size(Theta1,);
  size(Theta1,);

  for i=:size(Theta2,)
        for j=:size(Theta2,)
            JJ = JJ + Theta2(i,j)*Theta2(i,j)*lambda/(*m);
       end
 end
 size(Theta2,);
 size(Theta2,);
 %J = J + (lambda/(*m)) * (Theta1(:,:end).*Theta1(:,:end)+Theta2(:end,:).*Theta2(:end,:));
 J =J+JJ;

 Theta1_gradd = zeros(size(Theta1));
 Theta2_gradd = zeros(size(Theta2));

 for i=:size(Theta1,)
     for j=:size(Theta1,)
         Theta1_gradd(j,i) = Theta1(j,i)*lambda/m;
     end
 end

 for i=:size(Theta2,)
     for j=:size(Theta2,)
         Theta2_gradd(j,i) = Theta2(j,i)*lambda/m;
     end
 end

 Theta1_grad = Theta1_gradd+Theta1_grad;
 Theta2_grad = Theta2_gradd+Theta2_grad;

PS:博主蒟蒻强迫自己下次要写矩阵运算，不能再套循环啦！！！