“…7 is providing an empirical lower bound on the probability of identifying the global lower expectation. The implementation of simulated annealing follows [28] by using the parameters suggested in [29] and having 5 × |N k | moves per iteration, where |N k | is the size of the neighbourhood. The taboo search implementation is completely deterministic, the point in the neighbourhood that minimises E and is not on the taboo list is always chosen.…”