완료

ex4/nnCostFunction.m

@@ -39,6 +39,21 @@ Theta2_grad = zeros(size(Theta2));
 %         cost function computation is correct by verifying the cost
 %         computed in ex4.m
 %
+
+yy = zeros(m, num_labels);
+for i = 1:m
+  yy(i, y(i)) = 1;
+endfor
+
+a1 = [ones(m, 1), X];
+z2 = a1*Theta1';
+a2 = [ones(m, 1), sigmoid(z2)];
+z3 = a2*Theta2';
+a3 = hx = sigmoid(z3);
+J = sum(sum(-yy.*log(hx)-(1.-yy).*log(1.-hx)))/m;
+J += lambda*(sum(sum(Theta1(:, 2:end).^2)) + sum(sum(Theta2(:, 2:end).^2)))/(2*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
@@ -54,6 +69,15 @@ Theta2_grad = zeros(size(Theta2));
 %               over the training examples if you are implementing it for the 
 %               first time.
 %
+
+
+delta3 = a3 - yy;
+delta2 = delta3*Theta2.*[ones(m, 1), sigmoidGradient(z2)];
+delta2 = delta2(:, 2:end);
+
+Theta2_grad += (delta3'*a2)/m;
+Theta1_grad += (delta2'*a1)/m;
+
 % Part 3: Implement regularization with the cost function and gradients.
 %
 %         Hint: You can implement this around the code for
@@ -62,25 +86,11 @@ Theta2_grad = zeros(size(Theta2));
 %               and Theta2_grad from Part 2.
 %
 
+Theta2_grad(:, 2:end) += lambda*Theta2(:, 2:end)/m;
+Theta1_grad(:, 2:end) += lambda*Theta1(:, 2:end)/m;
 
 
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-% -------------------------------------------------------------
+% ----------------------------------------------------------
 
 % =========================================================================