Add regularization to the cost function

ex4/nnCostFunction.m

@@ -80,23 +80,14 @@ assert(size(J) == [1 1]);
 %               and Theta2_grad from Part 2.
 %
 
+% Note: Theta1/2 are matrixes here, we want all their rows, but skip their
+% first column (not regularizing the bias term).
+regularization_term = lambda/(2*m) * ...
+                      (sum(sum(Theta1(:,2:end).^2)) ...
+                       + sum(sum(Theta2(:,2:end).^2)));
+assert(size(regularization_term) == [1 1]);
 
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
+J += regularization_term;
 
 % -------------------------------------------------------------