Machine Learning ex4

function [J grad] = nnCostFunction(nn_params, ...
                                   input_layer_size, ...
                                   hidden_layer_size, ...
                                   num_labels, ...
                                   X, y, lambda)
%NNCOSTFUNCTION Implements the neural network cost function for a two layer
%neural network which performs classification
%   [J grad] = NNCOSTFUNCTON(nn_params, hidden_layer_size, num_labels, ...
%   X, y, lambda) computes the cost and gradient of the neural network. The
%   parameters for the neural network are "unrolled" into the vector
%   nn_params and need to be converted back into the weight matrices. 
% 
%   The returned parameter grad should be a "unrolled" vector of the
%   partial derivatives of the neural network.
%


% Reshape nn_params back into the parameters Theta1 and Theta2, the weight matrices
% for our 2 layer neural network
Theta1 = reshape(nn_params(1:hidden_layer_size * (input_layer_size + 1)), ...
                 hidden_layer_size, (input_layer_size + 1));


Theta2 = reshape(nn_params((1 + (hidden_layer_size * (input_layer_size + 1))):end), ...
                 num_labels, (hidden_layer_size + 1));


% Setup some useful variables
m = size(X, 1);
         
% You need to return the following variables correctly 
J = 0;
Theta1_grad = zeros(size(Theta1));
Theta2_grad = zeros(size(Theta2));


% ====================== YOUR CODE HERE ======================
% Instructions: You should complete the code by working through the
%               following parts.
%
% Part 1: Feedforward the neural network and return the cost in the
%         variable J. After implementing Part 1, you can verify that your
%         cost function computation is correct by verifying the cost
%         computed in ex4.m


% Part 2: Implement the backpropagation algorithm to compute the gradients
%         Theta1_grad and Theta2_grad. You should return the partial derivatives of
%         the cost function with respect to Theta1 and Theta2 in Theta1_grad and
%         Theta2_grad, respectively. After implementing Part 2, you can check
%         that your implementation is correct by running checkNNGradients
%
%         Note: The vector y passed into the function is a vector of labels
%               containing values from 1..K. You need to map this vector into a 
%               binary vector of 1's and 0's to be used with the neural network
%               cost function.
%
%         Hint: We recommend implementing backpropagation using a for-loop
%               over the training examples if you are implementing it for the 
%               first time.
%
% Part 3: Implement regularization with the cost function and gradients.
%
%         Hint: You can implement this around the code for
%               backpropagation. That is, you can compute the gradients for
%               the regularization separately and then add them to Theta1_grad
%               and Theta2_grad from Part 2.
%


K = num_labels;
X = [ones(m,1) X]; % 第一层 m*401
delta1_accum = zeros(size(Theta1));
delta2_accum = zeros(size(Theta2));
z_2 = X * Theta1'; % m*25
a_2 = sigmoid(z_2); % second layer m* hidden_layersize
z_3 = [ones(size(a_2,1),1),a_2] * Theta2'; 
a_3 = sigmoid(z_3); % third layer m*K
 
% J = -1/m * (y' * log(a_3) + (1-y')*log(1-a_3));  
% grand = 1/m * X' * (a_3 - y);  m*1


for i = 1:m
    y_newi = zeros(1,K); % initial y as a matrix
    y_newi(y(i)) = 1;    % when y = 5, set y(i) = 1
    J = J - 1/m * (sum(y_newi * log(a_3(i,:)')) + sum((1-y_newi)*log(1-a_3(i,:)')));
    
   delta_3 = a_3(i,:) - y_newi; % delta_3 is a 1*k
   delta_2 = delta_3 * Theta2 .* sigmoidGradient([1 z_2(i,:)]);  %delta_2 is 
   delta_2 = delta_2(2:end);
%    
    delta2_accum = delta2_accum + delta_3' * [1 a_2(i,:)];
    delta1_accum = delta1_accum + delta_2' * X(i,:);
end;


    Theta2_grad = 1/m * delta2_accum;
    Theta1_grad = 1/m * delta1_accum;
    
    J = J + lambda /(2*m) * (sum(sum(Theta2(:,2:hidden_layer_size+1).^2))+ ...
         sum(sum(Theta1(:,2:end).^2)));
    
    Theta2_grad(:,2:hidden_layer_size+1) = Theta2_grad(:,2:hidden_layer_size+1)+...
         lambda/m * Theta2(:,2:hidden_layer_size+1);
    
    Theta1_grad(:,2:end) = Theta1_grad(:,2:end) + lambda/m * Theta1(:,2:end);


% -------------------------------------------------------------


% =========================================================================


% Unroll gradients
grad = [Theta1_grad(:) ; Theta2_grad(:)];

end

应该是没什么问题了，还是要一步一步根据题目要求来检查才好发现问题，维度是最容易出问题的地方。