Sigmoid function in NN

X = [ones(m, 1) X];
temp = X * Theta1';

t = size(temp, 1);
temp = [ones(t, 1) temp];

h = temp * Theta2';
[max_num, p] = max(h, [], 2);

Without Sigmoid function, Training Set Accuracy: 69.620000

X = [ones(m, 1) X];
temp = X * Theta1';
temp = sigmoid(temp);

t = size(temp, 1);
temp = [ones(t, 1) temp];

h = temp * Theta2';
h = sigmoid(h);
[max_num, p] = max(h, [], 2);

With Sigmoid function, the Training Set Accuracy: 97.520000