nnCostFunction code

function [J grad] = nnCostFunction(nn_params, ...
                                   input_layer_size, ...
                                   hidden_layer_size, ...
                                   num_labels, ...
                                   X, y, lambda)

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));

m = size(X, 1);

J = 0;
Theta1_grad = zeros(size(Theta1));
Theta2_grad = zeros(size(Theta2));

X = [ones(m, 1) X];
A = sigmoid(Theta1 * X');
A = [ones(1, size(A, 2)); A];
t = sigmoid(Theta2*A);

J = (-1/m)*sum(sum(y'.*(log(t)) + ((1-y').*log(1-t)))) + (lambda/(2*m))*sum(sum(Theta1(:,2:end).^2)) + sum(sum(Theta2(:,2:end).^2));

for i = 1:m
    a1 = X(i,:);
    a2 = sigmoid(Theta1 * a1');
    a2 = [ones(1, size(a2, 2)); a2];
    a3 = sigmoid(Theta2*a2);
    delta3 = a3-y(i);
    delta2 = (Theta2'*delta3).*(a2.*(1-a2));
    Theta1_grad = Theta1_grad + [delta2(2:end,:)]*a1;
    Theta2_grad = Theta2_grad + delta3*a2';
end

Theta1_grad = (1/m)*Theta1_grad + (lambda/m)*[zeros(size(Theta1,1),1), Theta1(:,2:end)];
Theta2_grad = (1/m)*Theta2_grad + (lambda/m)*[zeros(size(Theta2,1),1), Theta2(:,2:end)];

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

end