Simulated annealing cooling schedule

Fig. 3.1. The cooling schedule during a simulated annealing run of 106 MC steps with goal curvature c0 = 10 in a box of unit edge length. The temperature program corresponds to a = 3. The adaptive changes in 6

It is probably correct to say that successful application of simulated annealing requires a clever definition ofndghborhoods in the configuration space (i.e., specification of individual "moves") and a ju dous variation of the control parameter p. In the case of metric spaces we would like to have a rationale for choosing an appropriate step size and for specifying a priori an effective cooling schedule. 

M. Huang, F. Romero, A. Sangiovanii-Vincentelli. An efficient general cooling schedule for simulated annealing. In Proc. of Int l Conf. Computer Aided Design, 1986, pp. 281 -384. 

In the last case where D > 0 and the probability is lower than the random value e d/T = Rnd, no moves are accepted and the current state s continues to be the current solution. When starting with a large cooling parameter, large deteriorations can be accepted. Then, as the temperature decreases, only small deteriorations are accepted until the temperature approaches zero when no deteriorations are accepted. Therefore, adequate temperature scheduling is important to optimise the search. Simulated annealing can be implemented to find a closest possible optimal value within a finite time where the cooling schedule can be specified by four components [2] ... 

M. T. Bakarat and P. M. Dean. /. Comput.-Aided Mol. Design, 4, 295 (1990). Molecular Structure Matching by Simulated Annealing. 1. A Comparison of Different Cooling Schedules. (And subsequent papers in this series). 

---