@staticmethod
    def calculate_cost(y, y_hat):
        num_train_examples = y_hat.shape[1]
        cost = np.sum(- np.dot(y_hat, np.log(y).T) - np.dot(1 - y_hat, np.log(1 - y).T)) / num_train_examples
        cost = np.squeeze(cost)

        return cost

    @staticmethod
    def __sigmoid(x):
        return 1.0 / (1.0 + np.exp(-x))

    @staticmethod
    def __sigmoid_derivative(x):
        return x * (1.0 - x)