1: \begin{abstract}

2:   We reconsider stochastic convergence analyses of particle swarm

3:   optimisation, and point out that previously obtained parameter

4:   conditions are not always sufficient to guarantee mean square

5:   convergence to a local optimum. We show that stagnation can in fact

6:   occur for non-trivial configurations in non-optimal parts of the

7:   search space, even for simple functions like \sphere. The

8:   convergence properties of the basic PSO may in these situations

9:   be detrimental to the goal of optimisation, to discover a

10:   sufficiently good solution within reasonable time.

11:   %

12:   To characterise optimisation ability of algorithms, we suggest the

13:   expected first hitting time (FHT), \ie, the time until a search

14:   point in the vicinity of the optimum is visited. It is shown that a

15:   basic PSO may have infinite expected FHT, while an algorithm

16:   introduced here, the Noisy PSO, has finite expected FHT on some

17:   functions.

18: \end{abstract}

19: