On the computational complexity of the probabilistic label tree algorithms