1: \begin{abstract}

2: We present the parallel and interacting stochastic approximation annealing

3: (PISAA) algorithm, a stochastic simulation procedure for global optimisation,

4: that extends and improves the stochastic approximation annealing (SAA)

5: by using population Monte Carlo ideas. The standard SAA algorithm

6: guarantees convergence to the global minimum when a square-root cooling

7: schedule is used; however the efficiency of its performance depends

8: crucially on its self-adjusting mechanism. Because its mechanism is

9: based on information obtained from only a single chain, SAA may present

10: slow convergence in complex optimisation problems. The proposed algorithm

11: involves simulating a population of SAA chains that interact each

12: other in a manner that ensures significant improvement of the self-adjusting

13: mechanism and better exploration of the sampling space. Central to

14: the proposed algorithm are the ideas of (i) recycling information

15: from the whole population of Markov chains to design a more accurate/stable

16: self-adjusting mechanism and (ii) incorporating more advanced proposals,

17: such as crossover operations, for the exploration of the sampling

18: space. PISAA presents a significantly improved performance in terms

19: of convergence. PISAA can be implemented in parallel computing environments

20: if available. We demonstrate the good performance of the proposed

21: algorithm on challenging applications including Bayesian network learning

22: and protein folding. Our numerical comparisons suggest that PISAA

23: outperforms the simulated annealing, stochastic approximation annealing,

24: and annealing evolutionary stochastic approximation Monte Carlo especially

25: in high dimensional or rugged scenarios.

26: \end{abstract}

27: